ISDG 2012 flame war

[Krill]

To bad i didnt posted, back time then but in pbem played against gaspar, lewwyn, seven.. I atacked lewwyn, and i lost i think , 6 batles at 78% knigs aginst LB, all after another, after i lost one batle at 96%, like after to win 3-4 at under 10% odds , one beeing even at under 4% i think - was a chariot against spear, is clear i was in tilt and very mad, about my luck, is this more probable to happen what happen for santa?
Against novice i won i think 6 batles at 50% odds knigs against pikes batles as well i lost 7 batles in row caravels against galeons, all having 33% , so i think this kind of streak can hapen just pray dont be on the wrong end.

Yeah, those are WAY more probable. Your caravel example happens about 1 in 17 times. The knights vs pikes is 1 in 64 times.

We are talking here about 1 in 44 billion times.

Several of the battles the whole statistical argument doesn't matter. For example, there is a 33% trireme v trireme battle in that sequence that was won. Was that cheating? Consider that in the context of thee game, if that trireme battle went ahead and was lost, the outcome was exactly the same as if the battle took place. Here is the turn report for that turn. Note that Kaz had naval numerical superiority in that area such that it would be possible for them to transfer units via galley to threaten the city and could stop the trireme from retreating to the city. Essentially our trireme had no value because it could not stop any attack from occuring, so we took a 1/3 chance of winning, the upside being a 4XP trireme that would have 50/50 odds on another battle after healing...and could retreat to an ocean tile so was safe from reprisal...or it died and we give Kaz trireme 1XP and that continue to threaten the city, status quo ante bellum.

Winning that battle is not evidence of cheating, but it contributes to the run of winning low odds battles. Yet making that attack is a valid tactic given the circumstances.

Another example of ignoring the potential outcomes is that according to Cyneheards combat spreadsheet (which I'm using now in lieu of the system Memphis was using because I no longer have access to that), the projected outcome of those battles has the defending units injured to a level where they would then be odds on battles to win on the second attack. eg:

T113: Features combat in defence of front city again Kaz maces. 1st attack: C2 axe v C1 Mace. Odds to win: 10.6%. Odds to drop mace to 52hp or lower: 59%. At that point, all SANCTA units would have greater than 50% odds to win. Kaz had 3 mace, 1HA, 1 chariot, 1 catapult versus 3 axes, 2 chariot, Archer, LB, spear and additional spear and LB to cover city for attack following turn.

The odds do not explain that enough defensive units in the city (Hello, CG2 LB here) the maces just needed to be damaged enough to not have odds on offence. For example a worse outcome on the first battle that lead to the mace being injured to 52 hp would still have been good enough to potentially wipe out that stack, but would definitely be good enough such that the attack on the front city could not occur the following turn. That enables further reinforcements to reach the city and defend. Like the artist to bomb it. So saying that because a win occurred it's cheating is just bullshit, when any of the cumulative outcomes with greater than 59% chance of occurring would have been enough to continue with the plan of attack.

[SevenSpirits]

The argument being made has nothing to do with whether the attacks were good or not. Sure, they were good attacks, it's not important.

The argument is: there are two ways you can get this lucky a result.
1) Through sheer luck. We expect this to happen less than 1 time ever (maybe way, way less).
2) Through cheating. We expect this to happen dozens or hundreds of times.

So of all the times that things like this have happened and will happen, probably 0-1 of them are legitimate and the rest are cheating. So we look at this case and in the absence of other information, ask what is the most likely explanation. In the absence of other relevant information, there's only one reasonable conclusion. It feels weird because you know there's a chance you're wrong just by random chance; you're very aware of that possibility. But there is always a chance you're wrong, though we rarely acknowledge it.

That's a nice segue though - I can't emphasize enough that it's easy to be wrong about your data or your math or your process. In my opinion, us knowing these numbers isn't enough to conclude anything, even if the result of the calculations says it's 99.99% sure.

[darrelljs]

If you agree with me that the odds are impossible

These are Bernoulli trials...all you can do is show how improbable the outcome is, because its never impossible.

I'm not sure I want to get further into it.

As an intellectual exercise it might be fun, but if the goal is to establish whether cheating occurred or not, your initial instinct was probably a good one .

Darrell

[mackoti]

I read all the battles and nothing seems wrong to me, So they actualy complain about 3 turns batles or just 1 turn?

[sunrise089]

(supposing you expect the a priori odds of a cheater to be one in a thousand, which is probably naive).

I'd guess this is generous by about two orders of magnitude

[Gaspar]

I wouldn't be able to complain about someone hitting 1 in 44 billion odds against me cheating or not because I would have already thrown my computer in a lake.

[Bobchillingworth]

This is an interesting discussion and all, but why is this thread being used to put a game from five years ago on trial?

[Commodore]

This is an interesting discussion and all, but why is this thread being used to put a game from five years ago on trial?

Because MJW is always on the side of the werewolves.

And yeah, BGN, that was kind of what I meant. The Sullla/Locke/Mikehendi/RBISDG Bog Standard Gameplan hasn't really changed. And drafting Maces is basically always a wake up call that you've really really lost your way.

I didn't pay much attention in the initial setup, but personally I was rooting against RB when Oledavy made that rapidly hushed-up mocking news article about the whatever guys' pick of that ginger chick.

[antisocialmunky]

I thought everyone but me forgot about that poorly thought out joke 0_o

[fluffyflyingpig]

You have to be careful when you're looking at data like this which is self-selecting because you wouldn't have looked at it if other teams didn't complain.

On that note, shouldn't some of the combat results leading up to SANCTA being accused of cheating be disregarded, as they are the reason you're looking at the data in the first place?

This reminded me of "The Science of Discworld" (of all books :lol:), which has a good and entertaining treatise on the use of statistics in science. You can preview every other page here:
http://books.google.no/books?id=Q1x0852i...&q&f=false
(It's surprisingly coherent even with every other page excluded). On page 273 they conclude: "It's illegitimate to include the data that brought the clump to our attention as part of the evidence that the same clump is unusual."

A valid point. As a purely intellectual exercise from a disinterested party, since this data set is timed you can test if an initial clump would raise a statistical flag and then look at subsequent combats. To whit:

SANCTA vs. Kaz (525BC)
axe v chariot, 4.5 v 4, 70.6%, win


SANCTA vs. Kaz (500BC)
axe v chariot, 4.29 v 3.82, 57.1%, win


SANCTA vs. Kaz (400BC)
axe v chariot, 4.52 v 4.40, 52%, win


SANCTA vs. Kaz (75BC)
trireme v trireme, 2 v 2.2, 32.2%, win


SANCTA vs. Kaz (50BC)
axe v mace, 6 v 8.8, 10.7%, win
chariot v mace, 4.4 v 8.8, 0.9%, lose (72/100 on mace)
archer v mace, 3 v 8, 0.2%, lose (52/100 on mace)
axe v horse archer, 5.5 v 6, 32.6%, win
axe v damaged mace, 5 v 6.3, 49.1%, win
longbow v catapult, 7.2 v 5, 78.3%, win
axe v chariot, 5.5 v 4.4, 75.1%, win
chariot v catapult, 4 v ?, 67.1%, win


SANCTA vs. Kaz (1AD)
chariot v war elephant, 4 v 8.8, 0.1%, withdraw (20% rate)
longbow v war elephant, 6 v 8.8, 13.5%, lose (36/100 on elephant)
chariot v maceman, 4 v 8.8, 0.1%, lose (100/100 on mace)
longbow v maceman, 6.6 v 8.8, 26.7%, win
spear v horse archer, 4 v 3.6, 73%, win
longbow v damaged war elephant, 6.6 v 3.2, 99%, win
axe v maceman, 5.5 v 8.8, 20.5%, lose
axe v damaged maceman, 6 v ?, 99.8%, win


SANCTA vs. Cavs (820AD)
knight v musket, 11 v 13.5, 25%, win
horse archer v musket, 6 v 13.5, 0.0%, withdraw (50% rate)
knight v musket, 11 v 13.5, 25%, win


SANCTA vs. Cavs (940AD)

cat v rifle, 5 v 24.5, 0.0%, lose (80/100 rifle)
cat v rifle, 5 v 19.6, 0.1%, lose
rifle v rifle, 15.4 v 22.4, 10.9%, win
knight v pike, 10 v 13.37, 21.7%, win
knight v pike, 11 v 13.37, 26.6%, win
rifle v knight, 14 v 10.23, 79.5%, win
knight v war elephant, 12 v 12.51, 47.5%, win
knight v pike, 11 v 12.05, 44.6%, win
spear v knight, 4.4 v 5.5, 24.7%, win
horse archer v war elephant, 6 v 11.42, 5.0%, withdrawl (30% rate)
knight v knight, 11 v 11.16, 49.3%, win
knight v knight, 12 v 10.32, 71.2%, win
knight v knight, 10 v 10.23, 48.8% win
axe v war elephant, 5 v 4.03, 85.2%, win

Test the combats up to 1AD. If this clump is improbable such to raise suspicion, discard that data and test the remaining data.


Leaving the actual calculation as an exercise for the reader (or someone with a few minutes to fiddle with mjw's spreadsheet), I'm going to go out on a limb and guess that this is all laughably unlikely.

Also, there is a wee bit of a difference between calculations saying you are four nines sure of something and calculations saying you are nine nine sure...