Slitherine

I think there might be a problem with your dices, gentlemen.

During this weekend I've played a lot of games, most of them in the v1.2.8, and there were some very strange things happening. I had lots of very weird cases of heavy armoured troops, both cavalry and foot, superior and drilled being broken and routed by fighting light foot missile troops. I thought it can't be just bad luck, as it was happening constantly, not only in a case or two.

So, I've started to follow the dices throw. After following them for a while, I think there might be some problem with the random generator you are using. Generally speaking, Attacker's dices are tending to be lower than Defender ones. I might be crazy, but you'd better have a look over this.

Bottom line is, things like my Marathon game from yesterday in which my hoplites line made contact without being shot at (therefore with steady troops) with the persians' mixed infantry line, and were almost all of them routed! and that without having a single persian unit routed in exchange. What are the odds for this to happen?!! There's definatelly something wrong with either the rules, or the random number generator (used for the dices).

It's possible but I'm pretty sure they're not. You have to check a lot of die rolls to get real data from this. I've played many many games and not noticed this as an issue.

You'd need to record it statiscially. Tabletop wargamers sometimes think their dice are broken but when peopel have recorded the stats they find they are actually producing average results but you don;t notice teh tthigns that go as expected or better, only the things that go against what you expect.

Exactly, the only ones which can do this are you guys, as we don't have access to the code to gather the thrown dices and make a statistic analysis on them. You can dump them from code and easly gather a few thousands of them to make a relevant comparison.

By the way, can we have the Summary (console) output of a game dumped into a file (with a time seed generated name, so it won't overwrite itself)?

But that would take a load of work and we have no reason to suspect they are wrong based on our experience

Hey Coth., the issue with marathon is that the Medium bow/spear Persians get 6 dice in the impact phase which will really rough up the Hoplites , if they are disordered they will lose badly in the melee phase as well
It really isnt a good idea to cahrge head long into the persian line in that scenario!

See http://www.hexwar.com/field-of-glory/he ... tcuts.aspx. For some keyboard shortcuts to change the display to show most die rolls.

Recording them would of course be an excercise in futility. You would need to roll in the order of of 100,000 die rolls to even begin to expect an even distribution and closer to 600,000. Within the game engine we do employ a doe roll tester that does roll 600,000 and guess what, it is 100% AOK.

Below that level you should expect and uneven distrubution. In fact if its even then there is more likely to be a problem. The numbers of should only even out if we were to compare the sum of the rolls of 10,000+ games where within each game the distrubution was probably uneven. Take a simple example.

Rolling 6 6's one after the other is the same odds's as 1 followed by a 2 then a 3,4,5,6. The simple mechanic applies at all levels. Roll 2 die with a 6 followed by a 6 is the same odds as a 6 followed by a 1 but the first seems lucky and the second fair. Roll these dice at the same time and the odd's change completely!

Sid Meier explains this well in http://uk.gamespot.com/pc/strategy/civi ... -sid-meier

So expect uneven die rolls, thats what the law of averages means for relativey small samples such as a few dozen games.
Keith

If you really want to know the maths behind the die rolls in the game, it's an additive congruential pseudo-random number generator.

http://comjnl.oxfordjournals.org/cgi/co ... t/11/3/341

Keith, I've studied Statistics in at the University, together with programming, mathematics and lots of other things (Economic Studies Academy university; Economic Cybernetics, Statistics and Infomatics faculty, Information Technology section 4.5 years).

I know what a Gaussian Distribution curve is. And you don't need 600.000 throws, and not even 100.000 to get relevant results for some six six-sided dices.

Also, I am always playing with Pre Combat Detail Display and Summary on and in verbose mode, always.

What I am telling you is that during this weekend I've got despaired regarding how my attacking units, with more POA than defending units, kept scoring 1s and 2s, even after re-rolling. I know that this can happen sometimes, even a few times during a match, even for a few games, but not in the amount it happened during this weekend. I've got cataphracts attacking light foot bowmen (1-1), and getting routed by those!

Maybe this was intended, I don't know in very great detail the FoG model, but if this is not a dice problem, then it is even more than a problem, it means the model is not right in representing some ancient warfare axiomata.

I'll dig for more info regarding the FoG combat model, but I'd say I have a strong feeling that dices are not so crystal clear at this moment.

Also, having Summary window's content dumped into a log file would be a great help into tracking the implemented fighting model and behavior.

PS: also, please do not understand I am saying that something it is broken in the random number generator. I am only saying that the probability of what happened this weekend is sooo low, that it should have not ever happened, and that the probability of something being not so quite ok in the random numbers generator is high enough that you might want to take a look at this, that's all

I think the point is that the die roller we use we test on 600,000 rolls and its AOK. The rest is just life! Keith

It would still be nice to be able to log the die rolls during a game, possibly split into a few categories like combat, cohesion tests and anarchy tests. I know that I some times end up wondering after a game whether I played something particularly badly or well, or if luck (either especially good or bad) was a big factor in how the game turned out. This would be less important if it was possible for a whole game could to be recorded for play back since that would make it easier to see the long term effects of particular decisions made during the game as well as how the dice contributed.

Chris