CompuServe Island Of Kesmia Reroller Analysis, 1991

From w.mfi
Jump to navigation Jump to search

CompuServe is almost completely dead now, but in the late 80s and early 90s it was one of the most popular "online" destinations. You called it over a (dedicated?) phone line with your modem at 300 or 1200 baud (bps) at a rate of $6.50 or $12.00 per hour respectively, I kid you not. (I really sank some money into them.) Later we moved to blazingly fast 2400 baud.

Island Of Kesmai was a text-based multiplayer exploration RPG. I only tried it briefly, but was intrigued by its reroller for "rolling your character" before playing. It simulates rolling a die; you would get X points for Strength, Y for Intelligence, etc.

So I wrote a little script that would roll it over and over and I let it run for hours while capturing the results. Then massaged the raw text capture (it's a text game sending ASCII characters over a modem) and put it into a dataset I could get statistics on.

My article was uploaded to the CompuServe MP Games IoK Techniques Library (library 7 out of 21 IIRC) on Jan. 24, 1991. Then I had the presence of mind to revisit CSi in late 1997 as I was moving away from it to copy down a few things, including this and the System Shock write-up.

Here's the analysis presented in its original form; only formatting was added. The actual pure text itself can be found here.

My "souped up 386" indeed. Time... what can you do but laugh. And cry.

Ref: An In-Depth Look at the IoK Reroller (REROLL.REF)

Submitted to MP Games (IoK/Techniques [7]) 
1/24/91 by 73717,250 - Zeus---*
Accesses: 143 as of 12/11/97   Size: 38743 Bytes

Description: The IoK reroller examined in relentless detail! Hours of fun for young and old alike. Version 3(27) examined. Enjoy!

Mike Fay


After seeing the low number of rolls in the existing reroller docs, and having a penchant for numbers, I decided to spend time rerolling and analyzing the results. I hope my efforts are of use to you.


Over 48,000 rolls were done, requiring many hours on the reroller at 2400 baud. Version 3(27) of IoK was examined (12/90-1/91). Rolls were captured, cleaned up, and fed into Paradox (a PC database) and Quattro Pro (a spreadsheet). Various analyses were done, such as comparing each of the six characteristics to Max Hit or Max Stam values. Thank heaven I had a souped-up 386 computer; some analyses took over an hour, and I did a lot of them.


1 = Illyria      ST = STrength
2 = Mu           DX = DeXterity
3 = Lemuria      IN = INtelligence
4 = Leng         WI = WIsdom
5 = Draznia      CO = COnstitution
6 = Hovath       CH = Charisma
7 = Mnar         MH = Max Hits
8 = Barbarian    MS = Max Stamina
B3 = the "basic three" characteristics (ST, DX & CO)
Bw = the "basic wiz" characteristics (B3 & IN)
Bt = the "basic thaum" characteristics (B3 & WI)
B6 = all six characteristics together (B3, IN, WI & CH)
n+g = nationality and gender (for brevity's sake)

Since Lynn de'Leslie states in STATS.REF that "wisdom has no relevance to any other character class" (besides thaums), it was not included in the Bw statistic. Data is provided below, however, if you want to include it in your definition of wizard. Also, if you are not familiar with the automated reroller program, look up the doc for IOKTERM in the MPGAMES forum.


Basic Observations

ST AND DX ADDS: You get one "add" for ST or DX if your ST or DX roll is 15 or higher. This is a well-established fact that I am repeating here simply for completeness. It was always found to be true in my data.

MS ADD: Stamina is linked to strength. If you get 15 or higher on your strength, you get MS 3 instead of MS 2. Period. This was an observation that fell right out of the analysis, but I can't find any indication that anyone else has noticed this. Most people I've spoken with thought Stam was related to CO or something. There may be some CO link when you increase in level, I don't know, but when rolling your character, there is none. The stam-strength link was always found to be true and will therefore not be discussed further in this doc. Getting a 15 for strength is therefore doubly important, since it gets you both a strength add and a stamina point.

MH: There does not appear to be any simple linkage of Maximum Hits to any other characteristic. I compared the MH roll to each of the six basic characteristics as well as MS for every n+g and could not find any correlation. It would appear to be simply the result of adding a six-sided die roll (1d6) to 20. For every n+g, MH always had a range of 21 to 26 and an average of 23.5 (3.5 is the average of six-sided die rolls). The greatest deviation from 23.5 for any n+g was seven one-hundredths (0.07). MH might be related to multiple other characteristics in a complex way, but, if it was, it probably would not have averaged 23.5 for each n+g; that would contradict the premise that it was in some way dependent on differences between nationalities and genders (unless all the differences were arranged to average 23.5 for each n+g!). I can only conclude that MH is a simple, random die toss (20+1d6), and that nationality, gender, and characteristic-rolls have no effect on your MH roll. Your chance of getting any particular MH value (21 to 26) is one in six. Therefore, I won't discuss MH "behavior" any further in this doc.

I recommend setting no target value for Hits on the reroller, since a few hit points have relatively little effect in the long run, whereas setting this to 26, for example, would make your reroll target take six times as long to generate, on the average. It's not worth 5 hit points.

Roll Stats By Nationality And Gender

Below is a listing of the average rolls for Nationality and Gender that I found, along with the number of rolls made for each.

Table 1. Average Rolls by Nationality and Gender.

N  S  AveST  AveDX  AveIN  AveWI  AveCO  AveCH   Num
1  M  11.07  12.05  12.11  12.01  12.05  12.01  3,332
1  F  10.01  12.07  13.07  12.14  11.97  12.03  2,373
2  M  13.99  12.01  10.00  13.12  12.04  11.03  3,499
2  F  12.99  11.93  10.07  13.08  11.87  11.06  2,352
3  M  12.97  11.98  11.89  12.12  10.96  12.09  2,316
3  F  12.04  11.92  12.14  11.97  11.04  12.05  2,347
4  M   9.94  12.93  13.14  10.93  10.01  14.05  2,336
4  F   9.98  14.04  13.02  11.09  10.03  14.04  4,250
5  M  12.04  12.06  14.89  12.05  12.14  11.04  8,487
5  F  11.04  13.02  13.94  12.10  12.05  11.14  2,335
6  M  10.98  11.02  10.97  13.89  13.01  11.97  2,333
6  F   9.82  11.00  11.13  13.91  13.99  12.11  2,304
7  M  13.98  11.98  11.09  11.02  13.04  12.04  2,598
7  F  13.02  12.14  11.10  11.05  13.08  11.95  2,323
8  M  11.92  12.08  12.05  12.06  12.04  12.10  2,325
8  F  12.00  12.06  12.02  12.02  12.10  12.01  2,522

N=Nationality (see abbreviations, above) and S=Sex (vive la difference!).

After having looked at these numbers and played with them in a spreadsheet for a while, it occurred to me that the average was generally 12 and that most numbers are close to a whole number. If you subtract 12 from each number and round to one decimal point (for clarity; see below), you get a table like this:

Table 2. Average Rolls with 12 subtracted.

N  S  AveST  AveDX  AveIN  AveWI  AveCO  AveCH
1  M  -0.93   0.05   0.11   0.01   0.05   0.01
1  F  -1.99   0.07   1.07   0.14  -0.03   0.03
2  M   1.99   0.01  -2.00   1.12   0.04  -0.97
2  F   0.99  -0.07  -1.93   1.08  -0.13  -0.94
3  M   0.97  -0.02  -0.11   0.12  -1.04   0.09
3  F   0.04  -0.08   0.14  -0.03  -0.96   0.05
4  M  -2.06   0.93   1.14  -1.07  -1.99   2.05
4  F  -2.02   2.04   1.02  -0.91  -1.97   2.04
5  M   0.04   0.06   2.89   0.05   0.14  -0.96
5  F  -0.96   1.02   1.94   0.10   0.05  -0.86
6  M  -1.02  -0.98  -1.03   1.89   1.01  -0.03
6  F  -2.18  -1.00  -0.87   1.91   1.99   0.11
7  M   1.98  -0.02  -0.91  -0.98   1.04   0.04
7  F   1.02   0.14  -0.90  -0.95   1.08  -0.05
8  M  -0.08   0.08   0.05   0.06   0.04   0.10
8  F   0.00   0.06   0.02   0.02   0.10   0.01

Notice that the farthest away one gets from an integer is 0.18 (6F/ST), and in general it is striking that the averages are so close to round numbers (for example, there are none around x.5, which would indicate a random distribution). If you round the numbers in Table 2, you get this:

Table 3. Roll Adds by Nationality and Gender.

  N  S  |  ST  DX  IN  WI  CO  CH  |  B3  Bw  Bt  B6 |
  1  M  |  -1   0   0   0   0   0  |  -1  -1  -1  -1 |
 Ill F  |  -2   0  +1   0   0   0  |  -2  -1  -2  -1 |
  2  M  |  +2   0  -2  +1   0  -1  |  +2   0  +3   0 |
 Mu  F  |  +1   0  -2  +1   0  -1  |  +1  -1  +2  -1 |
  3  M  |  +1   0   0   0  -1   0  |   0   0   0   0 |
 Lem F  |   0   0   0   0  -1   0  |  -1  -1  -1  -1 |
  4  M  |  -2  +1  +1  -1  -2  +2  |  -3  -2  -4  -1 |
 Len F  |  -2  +2  +1  -1  -2  +2  |  -2  -1  -3   0 |
  5  M  |   0   0  +3   0   0  -1  |   0  +3   0  +2 |
 Dra F  |  -1  +1  +2   0   0  -1  |   0  +2   0  +1 |
  6  M  |  -1  -1  -1  +2  +1   0  |  -1  -2  +1   0 |
 Hov F  |  -2  -1  -1  +2  +2   0  |  -1  -2  +1   0 |
  7  M  |  +2   0  -1  -1  +1   0  |  +3  +2  +2  +1 |
 Mna F  |  +1   0  -1  -1  +1   0  |  +2  +1  +1   0 |
  8  M  |   0   0   0   0   0   0  |   0   0   0   0 |
 Bar F  |   0   0   0   0   0   0  |   0   0   0   0 |

The summaries (B3, etc.) are presented for your convenience. These numbers seem rock solid, at least for the version of IoK tested, 3(27). Of course, only Kesmai Corp. knows if they will hold for future versions.

Once the above "adds" table became clear, I further analyzed the nature of the numbers contributing to the various adds. Specifically, I restructured my database and did an analysis of, for example, all the -2 adds. The following statistics were found:

Table 4. Grand Statistics by Adds.

 Add   Average   StdDev  SampleSize
  -2    9.9876   3.1052    23,700
  -1   11.0421   3.1107    55,038
   0   12.0414   3.0679   141,032
  +1   13.0402   2.9881    33,726
  +2   13.9913   2.8894    26,209
  +3   14.8914   2.6482     8,487  <-- Draznian(5) Male IN

StdDev is the standard deviation (of the sample, not the population, for you statisticians out there).

The distribution of die rolls for each add was found to be:

Table 5. Die Roll Percentage Distributions by Adds.

Roll     -2      -1       0      +1      +2      +3
   3   1.96%   0.84%   0.33%   0.09%   0.06%   0.01%
   4   2.03%   1.04%   0.47%   0.17%   0.07%   0.01%
   5   3.65%   2.14%   1.03%   0.50%   0.23%   0.02%
   6   6.03%   3.54%   2.00%   1.02%   0.48%   0.14%
   7   8.05%   5.40%   3.47%   1.87%   0.97%   0.37%
   8  10.20%   7.77%   5.50%   3.48%   1.93%   0.94%
   9  11.88%  10.27%   7.78%   5.43%   3.24%   1.78%
  10  12.79%  12.06%  10.16%   7.70%   5.30%   3.36%
  11  11.82%  12.46%  11.89%  10.12%   7.77%   5.40%
  12  10.57%  12.07%  12.76%  11.99%  10.21%   7.35%
  13   7.73%  10.57%  12.15%  12.48%  11.45%  10.04%
  14   5.73%   8.22%  10.59%  11.88%  12.30%  11.56%
  15   3.65%   5.80%   8.17%  10.72%  12.25%  12.95%
  16   2.14%   3.67%   5.77%   8.42%  10.80%  12.10%
  17   1.01%   2.18%   3.70%   5.87%   8.10%  10.83%
  18   0.77%   1.97%   4.24%   8.25%  14.82%  23.14%

Each column, above, adds up to 100%. Thus, for example, the Hovath(6) Female CO (+2) would be expected to get an 18 about 15% of the time, a 17 about 8% of the time, etc. If you multiply the die roll by its percentage occurrence and summate for all die rolls of a given add, you get the average in Table 4. Only the +3 adds have a small enough sample size that a graph of it shows some up-and-down "bouncing" along its points, indicative of a smaller sample size than optimal for an accurate picture, but even for it the bouncing is fairly minimal. (That's why I sampled the Draznian(5) male, the only configuration it appears in, more.)

A more useful presentation of the above tabulation would be in the form of how many times a given add would get a certain number or higher; for example, how often you would get a 15 or higher for a particular add. The following table simply summates the values from Table 5 for a given die value up to 18:

Table 6. Cumulative Die Roll Distributions by Adds.

Roll     -2      -1       0      +1      +2      +3
   3 100.00% 100.00% 100.00% 100.00% 100.00% 100.00%
   4  98.04%  99.16%  99.67%  99.91%  99.94%  99.99%
   5  96.02%  98.11%  99.21%  99.73%  99.87%  99.98%
   6  92.36%  95.97%  98.18%  99.23%  99.64%  99.95%
   7  86.34%  92.44%  96.18%  98.22%  99.15%  99.81%
   8  78.29%  87.04%  92.71%  96.35%  98.18%  99.45%
   9  68.09%  79.27%  87.21%  92.87%  96.25%  98.50%
  10  56.22%  69.00%  79.44%  87.43%  93.01%  96.72%
  11  43.43%  56.94%  69.28%  79.73%  87.71%  93.37%
  12  31.60%  44.48%  57.38%  69.61%  79.94%  87.97%
  13  21.03%  32.41%  44.62%  57.62%  69.73%  80.62%
  14  13.30%  21.84%  32.47%  45.14%  58.27%  70.58%
  15   7.57%  13.62%  21.88%  33.26%  45.97%  59.02%
  16   3.92%   7.82%  13.72%  22.54%  33.72%  46.07%
  17   1.78%   4.15%   7.94%  14.12%  22.92%  33.97%
  18   0.77%   1.97%   4.24%   8.25%  14.82%  23.14%

Thus, for example, the Leng(4) Male would be expected to roll 15 or higher on CO (a -2 add) about 8% of the time, a 16 or higher about 4% of the time, etc.


The Nature Of The Roll

I've already discussed how the Max Stam statistic appeared to be a simple case of adding a six-sided die roll to 20 (20+1d6). When I realized this, I analyzed the data for the different add values to see if the shape of their distribution could also be simulated by a die roll. Their distributions are extremely close to (but not exactly) what you would get by rolling three six-sided die (3d6). Rolling multiple dice gives you a bell-shaped curve that approximates what is known to statisticians as the normal distribution. The comparison is made awkward because the average roll of 3d6 is 10.5 (three times 3.5), and a simple graphing of 3d6 permutations gets you as many points at 10 as it does at 11. But, by doing some mathematical compensation to get the peak of the distribution centered on a whole number, you can get a curve that almost exactly matches the observed data. I tried a lot of other kinds of dice combinations (4d6, 5d5, etc.), some of which also came close, but 3d6 was the only six-sided die distribution that came close, and it was very close. Still, a real-live 3d6 with something added to it would never be able to get a 3 on the +3 adds (the curve would be shifted too far to the right to roll so low a value with three die), and this was observed in the data (once!). I can only imagine that the reroller uses some kind of computerized normal distribution function, supplying it with the average (12 in the case of 0 adds, for example) and standard deviation (approximately three, to simulate 3d6), and it spits back a value.

Values have a low of 3 and high of 18 cut-off point. Because of this, the -2 adds, with the midpoint of its curve at 10, is in the center of 3 and 18, and has the least amount of "tail crimping". As you go to the higher rolls, more and more values get cut-off at 18 and, although the reroller's probability function might have generated something higher, they become 18. A graph of the +3 adds has a big jag upward at the far right for the number of values at 17 and 18 (11% versus 23%, Table 5). Since values that would have been higher than 18 were reduced, this makes the over-all average for the +3 adds lower than the expected 15 (14.89, Table 4). Likewise, the other averages are not exactly on the mark (0 adds averaging 12.04 instead of 12, for example), despite that the reroller probably uses a bell-curve function centered on the average.

Planning Your Reroll

By using Table 6 (Cumulative Die Roll Distributions by Adds), you can predict how long it would take to generate a desired set of stats, on the average. For example, let's say you are interested in a wizard with a minimum of 15 ST, 15 DX, 18 IN, and 18 CO. You would see that the Draznian male had the following percent occurrence for these stats, respectively: 21.88% (Table 6: 0 adds, 15+), 21.88% (same), 23.14% (+3 adds, 18), and 4.24% (0 adds, 18). Multiplying these together gives you 0.04697%, or, its reciprocal, a 1 in 2,129 chance. In other words, you would expect to be able to roll the above-stated character in 2,129 rolls using a Draznian male, on the average. If you multiply this by however long it takes your CIS/IOKTERM reroller hookup to do each roll (get out your stopwatch!), you can estimate just how long it takes to get certain configurations. I usually get about 50 rolls a minute at 2400 baud, so it would take me about 43 minutes to get those stats, on the average. In reality, of course, it could take one roll or it might never happen; there are NO guarantees when it comes to probability.

Also, of course, note that your "winning" roll might actually have a stat better than 15 for something you'd targeted at 15, for example, so you may well get something better than your target. You can estimate the probability that you will get something higher than your target value beforehand. If you targeted 15 or higher for a +2 add, for example, the probability that you'll actually get an 18 is 32% (14.82%/45.97%, Table 6 +2 adds).

Optimal Characters Versus Nationality/Gender

Using the ideas presented in the previous section, you can analyze each of the nationality/gender combinations to find the one that best suits your desires. The following tables show the odds for selected target values, computed in the same way described above. For example, the Draznian(5) Male Wizard discussed in the previous section is seen in Table 8 under the "Bw" column (1 in 2,129 odds). These are configurations that, from my limited exposure to IoK, I think people would be interested in, and you can compute anything else you might want by combining Table 3 and Table 6.

"B3" and the other column headings refer to the basic stats players are most often concerned with (described in Abbreviations, above). Thus, B3 shows odds for fighters and MAs, Bw for wizzes, Bt for thaums, and B6 is all six, just for curiosity's sake. Refer to how I did the Draznian wizard, above, if you are confused.

Table 7. Odds of Getting 18 on all Relevant Statistics.

 N/S     B3          Bw          Bt          B6    
 1 M   28,236     665,942     665,942   370,428,672
 1 F   72,240     875,636   1,703,773   487,070,624
 2 M    3,753     487,449      45,495*  299,922,624
 2 F    6,742     875,636      81,726   538,770,048
 3 M   14,512     342,254     342,254   190,377,840
 3 F   28,236     665,942     665,942   370,428,608
 4 M  204,439   2,478,053  10,377,635   848,782,208
 4 F  113,807   1,379,483   5,777,024   472,500,192
 5 M   13,119      56,694*    309,412    67,874,560*
 5 F   14,512      97,919     342,254   117,228,632
 6 M   31,233   1,585,431     210,749   252,309,312
 6 F   44,483   2,258,025     300,156   359,347,520
 7 M    1,929*     97,919      97,919   117,228,632
 7 F    3,465     175,898     175,898   210,585,248
 8 M   13,119     309,412     309,412   172,109,520
 8 F   13,119     309,412     309,412   172,109,520
Table 8. Odds of getting at least 15 on ST, DX & CH and
          18 on IN, WI & CO.

 N/S       B3      Bw         Bt          B6   
 1 M      791    18,666     18,666    2,012,013
 1 F    1,424    17,260     33,583    1,860,475
 2 M      234    30,452      2,842*   2,710,133
 2 F      324    42,090      3,928    3,745,785
 3 M      698    16,451     16,451    1,773,313
 3 F    1,060    25,008     25,008    2,695,629
 4 M    5,158    62,523    261,833    6,903,931
 4 F    3,732    45,236    189,440    4,995,100
 5 M      493     2,129*    11,619      368,666*
 5 F      521     3,513     12,279      608,336
 6 M      653    33,168      4,409    1,022,893
 6 F      654    33,221      4,416    1,024,512
 7 M      121*    6,117      6,117    1,419,203
 7 F      167     8,455      8,455    1,961,538
 8 M      493    11,619     11,619    1,252,451
 8 F      493    11,619     11,619    1,252,451
Table 9. Odds of Getting at least 15 on all Relevant Statistics.

 N/S     B3      Bw     Bt       B6  
 1 M    153     701     701    14,642
 1 F    276     830   1,261    17,330
 2 M     45     600     137*   13,251
 2 F     63     830     189    18,314
 3 M    101     461     461     9,632
 3 F    153     701     701    14,642
 4 M    525   1,577   3,852    25,195
 4 F    380   1,141   2,787    18,229
 5 M     95     162*    436     5,428*
 5 F    101     219     461     7,365
 6 M    162   1,190     353    11,831
 6 F    211   1,549     459    15,401
 7 M     30*    219     219     7,365
 7 F     41     303     303    10,179
 8 M     95     436     436     9,114
 8 F     95     436     436     9,114

It would be almost impossible to list all permutations. You get the general idea. If you have a different target in mind, compute the odds by putting Tables 3 and 6 together.

Gender Comparison (My Favorite Kind!)

By looking at the adds table (Table 3), you can see that adds are much more consistent within a nation than they are between nations. Gender never makes more than one add-point difference in a given characteristic within a nation, and usually there's less than two characteristics changed between male and female (sounds like ... er, skip it; bad joke). Thus, gender consideration is less important than nation consideration. Past this general observation, there are some things to note.

In general, adds for females are worse than for males. (Don't look at me; I didn't design the reroller!) It's not that fems don't have any advantages; they just don't seem to amount to anything really useful.

The biggest difference is that six out of eight nations have one less strength point for the female. Since strength is a basic characteristic that most any character is interested in, fems are pretty much behind the eight ball due to this alone. Past this, there are five other gender differences. The Illyrian(1) female has an extra IN point, but coupled with a poor -2 strength, it's a bad bet. Leng(4) fems have a DX add, but Leng already has horrible ST and CO, making it a bad nation in general. The Draznian(5) fem has a DX plus, but Draznia is most noted for its IN, and the fem has a reduction there. Finally, the Hovath(6) fem has a CO plus, but a ST neg (for a whopping -2), with this on top of generally poor nation abilities in other areas. Generally a lack-luster showing, possibly leaving males as the optimal choice for all character types.

You might have thought that fems could have been designed to give a decisive advantage somewhere or other. Maybe the designers thought that IoK was mainly about combat, that males are usually more interested in fighting (this is not necessarily good! <grin>), and that therefore fems had to try a little harder to survive. Nevertheless, it's not as bad as it may appear if you want a female character. The reroller is forgiving, and one less add-point is not the end of the world. A straight-18 male Mnar(7) fighter, for example, has roll odds of 1,929 (Table 7, B3), whereas the female has 3,465. At my average 2400 baud reroll rate of 50 per minute, this means 39 minutes for the male versus 69 for the female, little difference for the dedicated IoKer.

News Flash: Just prior to uploading this document, an IoK person said that they had also noticed the problem with females stats, and that they might do something about it. It was a spur-of-the-moment, off-the-record comment, though. Therefore, females/the adds table (Table 3) might be adjusted for this sometime in the future. If you think you see a difference from my numbers indicating the adds table has changed, you are welcome to record your rerolling, zip the file, and send it to me, at least at the current time (early 1991). The IOKTERM capture buffer works just fine for capturing rerolls. If I am still caring about this <grin>, I'll look into it and possibly update this document. Make sure you capture which nation and sex you were rolling, or that you otherwise tell me. Normally, it is only captured once at the very beginning of rerolling; the "print screen to disk" function does not print into the capture buffer.

Adds By Nationality And Comparison With Their Description

Let's take a look at the adds given to nations (Table 3), and kill two birds by also comparing the description of the nations found in IoK's main menu document area ("Describe Nationalities" on the IoK menu; "Nationalities on Kesmai" at the beginning of its text). For brevity's sake, I'll sometimes refer to nations by using Nx (e.g., Illyria(1) = N1). Quoted sections are paraphrased from the "Nationalities on Kesmai" text. NOTE: These are paraphrases, not direct quotes, since the description is sometimes wordy plus I'm always lazy <grin>.

As we'll see in the comparisons that follow, the description saying a nation is good in something usually really just means it's not bad at it, and only half the time when a special "plug" is put in for a particular characteristic does the nation actually have a plus in that category. Kesmai Corporation has made a good choice in arranging the wording in this "soft" way, because the alternatives would have been either to say almost nothing concerning aptitudes, or to give everything away.

Gender differences have been described already, so I will only discuss the general nation differences in this section.

ILLYRIA(1): Some relevant paraphrases are:

"N1 nobility claim their descent from the Elves of old",
"N1s as a rule are competent fighters",
"Wizardry is also common since N1s tend to be practical people and quite distant from the hotbed of Thaumaturgy in Hovath",
"The Imperial Institute of Wizardry is perhaps the finest in the world",
"Sorcery is completely outlawed"

N1 is -1 in ST and otherwise flat. I guess their elven descent makes them a wee bit tiny, but they are otherwise generally all-around capable. They don't have an IN add due to Kesmai Corp.'s apparent "soft" policy (see above).


"Mu is the home of the Grand Order of Thaumaturges",
"Citizens of Mu tend to be stocky, heavy-set, and even the women are more muscular than average",
"Thaumaturges are common among wayfarers from Mu, as are fighters. Thieves are not unheard of, but wizards are rare, and [sorcery outlawed]",
"A muscular Muian fighter is considered deadly by all other races"

Mu has a healthy ST plus, a WI plus, a CH impairment, and a severe IN impairment. This is very consistent with being a muscular race good at thaumaturgy which frowns on wizardry and sorcery. (For those of you not aware of it, Lynn's STATS.REF states that higher CH will make you a better sorcerer, if sorcery is ever implemented in the game. Currently, there is no sorcery, and CH has no function.) The selected odds tables shows the Mu male as being the best thaum, including a +2 ST that will make it easy to give him ST 18, if you want that for your thaum.


"They are basically a good people, but in a low key way... whose basic philosophy of governing is to do as little as possible",
"Wizards, fighters, and thaumaturges are common, but thieves are rare, and sorcerers unheard of",
"Their Knights are proficient at fighting",
"N3s are singularly unimpressed with the High Priest of Hovath",
"No particular racial characteristics predominate",
"N3s love bright colors... and bright gem stones"

Lemuria is -1 CO and otherwise flat. The adds don't say you can't be a sorcerer from N3 (although they also don't discourage it), so I guess that's a cultural and not a genetic barrier <grin>. With the good-naturedness of N3s, love of bright colors, and low CO, maybe they're just the prototypical laid-back crowd of KesmaiWorld (kind of like Californians?).


"The Guild Of Sorcerers is strong in Leng",
"Moral concerns are not very important in Leng",
"The people of Leng are smaller than usual, but very quick",
"Much of the life of Leng is carried out in the early evening and night",
"It is said that a strange race came to Leng and interbred",
"The Lesser Ring is composed of elite fighters",
"Sorcerers, fighters and thieves are numerous, and wizards not too uncommon, but thaumaturges are unknown",
"Their fighters more than make up for their lack of size by their speed, training, and bloodlust"

Leng is the mishmash of the Adds Table, with values ranging from -2 to +2 and nary a zero to be found. I guess the race they interbred with was strange indeed. As befits the description, ST is very bad and DX is good; CO, however, is also very bad. Likewise, IN is good, but WI bad. CH is the highest of any nation, a +2 for both male and female, so it may be good for sorcery some day if you can swallow the low ST and IN.


"The green color of the island is echoed by the clothing of the people",
"There is much magic in Draznia",
"N5 has produced many, if not most, of the great wizards of history, and is likely to produce more",
"A typical Draznian will be of average height and build",
"Wizards are very common, as are fighters; there are a few thaumaturges and some thieves, but rarely any sorcerers"

N5 is far and away the best IN nation (+3 for male). This is the only +3 in the whole Adds Table. Most of the rest is flat, although there is a negative CH working against would-be sorcerers. It is the best in the selected odds tables for wizardry. (The description of N5s sounds kind of like Irish leprechauns, begorrah!)


"The High Priest of N6 is the supreme authority ... with harsh penalties for wizards",
"Thaumaturges are supreme in N6",
"Wizards and sorcerers are heretics to be burned at the stake",
"N6s are tall and strong, but only average when it comes to quickness",
"Thaumaturges are common, outnumbered only by fighters, and there are also many thieves",
"The few wizards and sorcerers are still living only because they are well hidden",
"The climate is hot, and the people spend the middle part of the day napping",
"There is much dissatisfaction in N6, but it is totally disorganized",
"There is a cadre of powerful assassins"

N6 has a whopping +2 WI as well as a plus for CO, with CH flat and ST, DX and IN negative. The claim for thaumaturgy is clearly supported, but the negative B3 stats drag down the odds for a thaumaturge, making Mu(2) surpass it. While the positive CO helps fighters, the negative ST and DX detract from it, making them not too good for fighting, either. Obviously there is a pitch for thieves from N6 as well. I don't know too much about thieves, but I can only imagine that DX is very important to them, so the negative DX would hamper being a thief from Hovath, and the negative IN does not help either. The negative IN and flat CH is also not encouraging for wizards or sorcerers.


"N7 is a land of roving herders",
"Their people are tall and strong",
"They are noted warriors",
"Most N7s are fighters, but wizards, thaumaturges, sorcerers and thieves are also represented; thieves can gain much honor in Mnar, and sorcerers are tolerated"

Mnar has good ST and CO bonuses, but IN and WI negatives, with the rest being flat. It is best for fighters in the selected odds tables by a country mile. The implication that there are not many wizards, sorcerers, or thaums from N7 is supported by the low IN and WI adds.


"N8s wander on the wide plains and deserts of the continent",
"They are big-boned and strong; not overly tall, and their wide shoulders and thick necks make them appear even shorter than they are",
"All the tribes produce fighters",
"To be considered a chief, N8s would have to be a great fighter or thaumaturge",
"No thief would be considered a chief unless also proven a fighter or thaumaturge" (Thief the Chief?),
"The Shriker clan has many great thaumaturges",
"The Wind Knot clan consists mostly of fighters",
"The Grass Snake clan has many thaumaturges and even more who are thieves",
"The Mountain Cloud clan is the only one to have any wizards",
"The Dark Moon clan is famous for producing sorcerers"

With a surprising twist, Barbarians are totally flat for both males and fems. Thus, they are not particularly good or bad for anything, and the text implies that all classes are possible with them. This adds support to my thought that Kesmai has given a "soft" description of nations' abilities, so that people would neither feel compelled to pick one nation for something nor not have any idea what to pick.

Summary Of Text Versus My Data

If you accept the premise that Kesmai did not want to force people into taking certain nations for a given type of individual, then my findings are, on the whole, in very good agreement with the text description. While there are some possible contradictions, the text is usually pretty mushy when it comes to saying exactly what a nation can do anyway, so this is not a big concern. And almost always when my data said a nation was very good or very bad at something, the text supported it.

Class Versus Nationality Recommendations

Given the selected odds tables and the text description, it seems logical to assume the following: A Mnar(7) male is best for fighting, a Draznian(5) male is best for wizardry, and a Muian(2) male is best for thaumaturgy. I'm not sure just what an MA needs, but I gather it's the same as a fighter.

Thieves are a mixed bag. For one thing, I don't know much about them and can't say as I'm qualified to know what stats they need best. I would think DX, CO, and IN would be the primary ones. If this is true, one would have hoped that the Leng(4) female would have been a good choice (DX +2), but the horrible CO scores (-2 for both male and fem) make it almost impossible to get CO 18 with a Lengian. Almost no other nation has a DX add. Past that, maybe a Mnarian(7) male or Draznian(5) female would be best. It's definitely a case where specific goals need to be known, and then Table 6 (Cumulative Die Rolls) consulted.

Since I'm not sure what a sorcerer would need, if it ever is introduced, it's hard to say which nation would be best. But: If it needed a good IN, CH, and B3, Illyrian(1) females or Barbarians(8) might be a choice. I say this not because they are good in the relevant stats, but because they are not bad in anything needed. Since those five stats involve everything but WI, it's a very tough roll and involved decision, hardly worth worrying about before even knowing what's actually needed. 'Nuff said.

For what it's worth, the Leng(4) male seems to be the worst for almost anything.

Using The Reroller

If you watch your statistics as they flash up on the screen of the IOKTERM reroller, you will see some interesting things happen relative to my data. For one thing, I have found that the average ("mean") shown on the reroller screen is often way off until you start getting to about 200 rolls. Remember, I've composed big tables of averages here; the actual numbers you get might be as fickle as mail delivery until you start getting the big sample sizes I had. A standard deviation of three (Table 4) is nothing to sneeze at. Often, though, once you have gotten past this point, your medians will be right on the mark, even if the averages are off by 0.4-0.6 or whatever, slowly closing in on mine.

If you are trying for an 18 that is hard to get (Add -1, for example), your "percent hit" number will often be quite different from mine (Table 6) until you get a quite large sample size (400 or even 1,000). This is because if you've only gotten one "hit" on a toughie, for example, a second hit will double the percent hits on the reroller screen, leading to plenty of bouncing around.

After you have a large sample size, your odds rating will start to approximate the odds you estimate from my data. Again, though, setting a characteristic to an improbable value will cause sizable swings in this value, depending on how lucky (or unlucky) you are.

Conclusion / Author's Notes

I don't know if it's only because I'm new to IoK so I haven't been around to hear what others had to say on the reroller, but I am surprised that at least some of the analysis I performed has not been figured out already. Lots of people who are good with computers must have been through IoK, and the reroller is an eminent example of numbers waiting to be crunched. The stamina/strength link seems clear and the adds table fairly obvious, although the underlying 3d6 function might have eluded discovery. Then again, I have run into a lot of Mnarian fighters and Draznian wizards out there and the existing reroller docs lean toward my findings, so people have definitely picked up on some of this. I guess that since players were (1) paying for the time involved, and (2) wanting to get on with the game, they just got rerolling over with and got on with the game.

Kesmai Corporation may some day change their reroller. Unless they change the basic 3d6 nature of the reroller, however, you will quickly be able to generate a new picture of the lay of the reroller, armed with a knowledge of its general features as shown in this article. For example, if they switch the adds for females around, the same cumulative distributions may apply (i.e., Table 6 stays the same), but the nations and genders would have different add configurations (i.e., Table 3 changes).

In truth, I can't see them changing the underlying 3d6 function much, because it is a very good simulation of the kind of distribution they probably want, and anything else might make good characters either too easy or too hard to get. Plus, they might have to rewrite the nation descriptions if they change the reroller. In fact, the automated IOKTERM reroller itself has been more of an assault on making good characters easy than this document because, more important than what you know, is the time you are willing to put into rolling good stats. The automated reroller blows away anything you could do manually.

One more thing to keep in mind is that my statistics are a simple tabulation of reroller numbers. If anything else is going on that those numbers won't reflect, like secret weapons, magic, or luck modifiers, it has been totally missed in this document. I can't recall coming across anything about this in the MPGAMES Libraries, though.

In closing, I can't emphasize strongly enough that a standard deviation of three (which all characteristics use) means that almost anything can be achieved, if you have enough patience.

       Keep your spells warm and your shields up!

                       Mike Fay
                    January, 1991