Post by truealcove on Mar 19, 2018 19:23:17 GMT -5
(Sorry, when I copy and pasted I did it poorly the first time. Now this post should look right.)
So I decide to test today a few things that I was suspecting before, and found some possible bugs. Let me first state my observations, and say why I think it's the case.
1. Strong dices seems to have a 40% success rate instead, and standard dice 20% (doesn't quite affect the game as it applies to both templars and enemies anyway).
2. Shield and Warrior parry bonus seems not working as indicated; Warrior looks like giving standard dices, and shield parry seems close to not existing.
3. Grenades critical damage bonus counts only Focus, but neither Willpower nor Grenades.
4. The formula for "Chance to Hit" indicator seems to wrong.
Probably 2 is the only one that matters to most people. But I suppose 1~3 (if there are really bugs) might have easy fixes, so I'll bring them up at the same time
----
So for 1 and 2, I get the test data from gamelog. It might be that gamelog is where the bug is. But if the gamelog is buggy in this implicit manner (that is, the # of dices are correct but the # of success rolls is not), then we should see something like attack 6 successes vs defense 7 successes yet still damage done? So I prefer to guess the bug is not from gamelog.
1. Strong dices and standard dices.
I have three test data:
(A) Produced from 54 templar ranged attacks. 494 strong dices and 816 standard dices are rolled. Get 364 successes total. Checked in advance that the # of strong dices and standard dices in the combat log is the same as indicated by game rules.
(B) Produced from 53 xeno dodge. 165 strong dices and 198 standard dices are rolled. Get 95 successes total.
(C) Produced from 55 xeno melee attacks. 330 strong dices and 1923 standard dices. Get 511 successes total.
If strong dices have 50% success rate and standard dices have 30% success rate, then in (A) we expect 494*0.5+816*0.3=491.8 successes. In (B) we expect 141.9, and in (C) we expect 741.9 successes. All actual outcome are much higher than expected. If we sum them all, we have 989 strong dices and 2937 standard dices which produce 970 successes, while 989*0.5+2937*0.3=1375.6.
If we roll a dice N times, the variance is at most N*0.25 (at most 0.25 for any kind of success/fail dice, coin, etc). So in total the variance is no more than 1000, and the standard deviation is the square root of that, which is no more than 32. You can't really have results that's off from expected value by 5 times standard deviation or so; that happens only once a million times (or Wikipedia says so)
Since the dice is said to be 1D10, I'd guess that it's 40%/20% each instead; in which case 989*0.4+2937.0.2=983, close enough to the outcome 970. Maybe a >= is mistyped as a >, or floating number error?
----
2. Shield and Strength for parry
So I tested three groups of data from the gamelog.
(A) Captain with Blade (3s), Orion Shield (8s), 16 Warrior(16s) and 3 Evasion (3) = 27s+3. Tested 56 times. The average number of success rolls is 5.11.
(B) Captain with Blade (3s), some sidearm (0), 4 Warrior(4s) and 15 Evasion (15) = 7s+15. Tested 54 times. The average number of success rolls is 6.24
(C) Paladin with Heavy Warblade (4s), Orion Shield (8s), 14 Warrior(14s) and 3 Evasion (3) = 26s+3. Tested 53 times. The average number of success rolls is 6.58.
I actually can't have a theory of how the result of (C) and (B) are higher than that of (A). But the general idea is that Shield evidently doesn't work, and Warrior is not better than Evasion for melee defense. For reference, the standard deviation for throwing 30 coins/dices 50 times and compute the # of success is no more than 0.5*sqrt(30/50)~0.39.
----
3. In imgur.com/a/drHDZ, we have a captain using level 1 grenade (47-86) with 2 Focus, 10 Willpower and 3 Grenade. The damage is 29=53-24 (the xeno actually lose 29 life, so it's correct for our interest). As 53 is just too low (47*1.18=55.46), this can just mean that only Focus boosts grenade critical damage.
----
4. This is a tricky one. I bet I am not the only person who boosts accuracy a lot and wonder why my "Chance to Hit" improves so little. Eventually I write a program myself and get some conflicting result. For example, in imgur.com/a/ARMN6 the chance to hit is shown to be 64%, while 9s+6 vs 4s+5 will in fact give about 90%. (I also tested it 25 times and get 24 hits and 1 miss).
So I decide to test today a few things that I was suspecting before, and found some possible bugs. Let me first state my observations, and say why I think it's the case.
1. Strong dices seems to have a 40% success rate instead, and standard dice 20% (doesn't quite affect the game as it applies to both templars and enemies anyway).
2. Shield and Warrior parry bonus seems not working as indicated; Warrior looks like giving standard dices, and shield parry seems close to not existing.
3. Grenades critical damage bonus counts only Focus, but neither Willpower nor Grenades.
4. The formula for "Chance to Hit" indicator seems to wrong.
Probably 2 is the only one that matters to most people. But I suppose 1~3 (if there are really bugs) might have easy fixes, so I'll bring them up at the same time
----
So for 1 and 2, I get the test data from gamelog. It might be that gamelog is where the bug is. But if the gamelog is buggy in this implicit manner (that is, the # of dices are correct but the # of success rolls is not), then we should see something like attack 6 successes vs defense 7 successes yet still damage done? So I prefer to guess the bug is not from gamelog.
1. Strong dices and standard dices.
I have three test data:
(A) Produced from 54 templar ranged attacks. 494 strong dices and 816 standard dices are rolled. Get 364 successes total. Checked in advance that the # of strong dices and standard dices in the combat log is the same as indicated by game rules.
(B) Produced from 53 xeno dodge. 165 strong dices and 198 standard dices are rolled. Get 95 successes total.
(C) Produced from 55 xeno melee attacks. 330 strong dices and 1923 standard dices. Get 511 successes total.
If strong dices have 50% success rate and standard dices have 30% success rate, then in (A) we expect 494*0.5+816*0.3=491.8 successes. In (B) we expect 141.9, and in (C) we expect 741.9 successes. All actual outcome are much higher than expected. If we sum them all, we have 989 strong dices and 2937 standard dices which produce 970 successes, while 989*0.5+2937*0.3=1375.6.
If we roll a dice N times, the variance is at most N*0.25 (at most 0.25 for any kind of success/fail dice, coin, etc). So in total the variance is no more than 1000, and the standard deviation is the square root of that, which is no more than 32. You can't really have results that's off from expected value by 5 times standard deviation or so; that happens only once a million times (or Wikipedia says so
)Since the dice is said to be 1D10, I'd guess that it's 40%/20% each instead; in which case 989*0.4+2937.0.2=983, close enough to the outcome 970. Maybe a >= is mistyped as a >, or floating number error?
----
2. Shield and Strength for parry
So I tested three groups of data from the gamelog.
(A) Captain with Blade (3s), Orion Shield (8s), 16 Warrior(16s) and 3 Evasion (3) = 27s+3. Tested 56 times. The average number of success rolls is 5.11.
(B) Captain with Blade (3s), some sidearm (0), 4 Warrior(4s) and 15 Evasion (15) = 7s+15. Tested 54 times. The average number of success rolls is 6.24
(C) Paladin with Heavy Warblade (4s), Orion Shield (8s), 14 Warrior(14s) and 3 Evasion (3) = 26s+3. Tested 53 times. The average number of success rolls is 6.58.
I actually can't have a theory of how the result of (C) and (B) are higher than that of (A). But the general idea is that Shield evidently doesn't work, and Warrior is not better than Evasion for melee defense. For reference, the standard deviation for throwing 30 coins/dices 50 times and compute the # of success is no more than 0.5*sqrt(30/50)~0.39.
----
3. In imgur.com/a/drHDZ, we have a captain using level 1 grenade (47-86) with 2 Focus, 10 Willpower and 3 Grenade. The damage is 29=53-24 (the xeno actually lose 29 life, so it's correct for our interest). As 53 is just too low (47*1.18=55.46), this can just mean that only Focus boosts grenade critical damage.
----
4. This is a tricky one. I bet I am not the only person who boosts accuracy a lot and wonder why my "Chance to Hit" improves so little. Eventually I write a program myself and get some conflicting result. For example, in imgur.com/a/ARMN6 the chance to hit is shown to be 64%, while 9s+6 vs 4s+5 will in fact give about 90%. (I also tested it 25 times and get 24 hits and 1 miss).
