Project ELO (Dataset released)

[mhsmith0]

Oh. I'm not doing the same thing as realeo, so my rules and his don't overlap.

[Loopdan]

A bit of a flip from 1.001 - Rask's rating took a hit from the loss, and Loop's rating got a bump from the 1767 win

take that Rask!

[mhsmith0]

For kicks I tried looking at my own town rating (on a 2-2 record) - that 1756 loss was CRUSHING. JaeReed and Loop both high score villagers, and I got rated at a loss w/ them, despite GOATing like crazy there.



Also I lost 1691 with Thor, another unfortunate result that doesn't really help.

Guess I need to start winning more or something

[Raskolnikov]

Soon.

[mhsmith0]

lol went back to research, 1370 actually was the first ever matrix6 http://wiki.mafiascum.net/index.php?title=Matrix6
so my stats are all matrix6 games, fun!

[Realeo]

Repository updated.

[Realeo]

I am confuzed.

I notice there is hydra when compiling Open. Should I split the hydra or not?

[callforjudgement]

IMO treat the hydra as an account on its own. Players play differently inside and outside hydras.

[mhsmith0]

Update for 1768 and 1770, plus a couple other fixes:

Spoiler: v1.003 results

	Log Rating			Record
Player	Town	Scum	Combined	Town	Scum	Combined
GuyInFreezer	1.64		1.64	12-3	1-1	13-4
Thor665	0.46	-0.40	0.86	12-13	6-3	18-16
Malakittens	0.85		0.85	17-10	3-3	20-13
fferyllt	0.76		0.76	12-5	3-1	15-6
T S O	0.71		0.71	7-5	2-2	9-7
Xayzeck	0.58		0.58	7-4	1-4	8-8
Micc	0.57		0.57	5-4	2-1	7-5
singersigner	0.55		0.55	6-5	1-0	7-5
Huntress	0.55		0.55	7-5	1-3	8-8
Drixx	0.48		0.48	11-10	3-2	14-12
JaeReed	0.39		0.39	5-4	0-0	5-4
RadiantCowbells	0.34		0.34	14-15	4-1	18-16
Hopkirk	0.33		0.33	5-3	2-0	7-3
Bulbazak	0.28		0.28	5-4	2-2	7-6
Loopdan	0.26		0.26	5-5	2-0	7-5
Guyett	0.26		0.26	7-7	2-2	9-9
Raskolnikov	0.23		0.23	5-6	1-2	6-8
tojam2	0.17		0.17	3-5	0-1	3-6
theslimer3	0.09		0.09	4-4	0-0	4-4
jmo16mla	0.08		0.08	5-5	0-0	5-5
hayatoBL	0.03		0.03	5-6	3-1	8-7
TheIrishPope	0.00		0.00	4-5	1-3	5-8
innocentvillager	-0.02		-0.02	5-7	4-1	9-8
Wisdom		0.04	-0.04	3-4	6-4	9-8
copper223	-0.10		-0.10	3-5	2-0	5-5
goodmorning	-0.20	-0.09	-0.11	14-23	7-5	21-28
Nachomamma8	0.70	0.83	-0.13	17-12	5-12	22-24
House	-0.18		-0.18	5-7	2-0	7-7
Nobody Special	-0.19		-0.19	6-8	1-2	7-10
JasonWazza	-0.19		-0.19	4-5	3-3	7-8
RachMarie	-0.19		-0.19	11-18	2-2	13-20
Dierfire	-0.21		-0.21	6-10	2-2	8-12
Sakura Hana	-0.22		-0.22	5-6	3-3	8-9
Accountant	-0.23		-0.23	6-11	2-1	8-12
Not_Mafia	-0.30		-0.30	6-10	3-4	9-14
PhantomCobalt	-0.32		-0.32	5-8	1-1	6-9
BlueBloodedToffee	-0.13	0.21	-0.34	10-17	6-5	16-22
Ms Marangal	-0.41		-0.41	3-6	1-4	4-10
NicCage	-0.45		-0.45	2-6	0-0	2-6
Cabd	-0.46		-0.46	6-8	0-0	6-8
notscience	-0.29	0.19	-0.48	9-16	5-4	14-20
Alexcellent	-0.49		-0.49	3-5	0-1	3-6
enomis	-0.51		-0.51	2-6	0-0	2-6
Flubbernugget	-0.70		-0.70	3-8	0-0	3-8
MarioManiac4	-0.87		-0.87	2-8	1-2	3-10
Jake from State Farm	-0.91		-0.91	2-10	1-3	3-13
Titus	-1.00		-1.00	3-12	0-2	3-14

intercept: -0.403 (40.1% town win odds given entirely "other" players)
combined error: 248.87 (average 0.624)

Spoiler: updates and corrections list

Added newbies 1768 and 1770
RC added to newbie 1711
Added tojam2 to tracking list (has now reached 8 town games)

Interesting changes and other notes:
Nothing in particular from this update tbh

[LicketyQuickety]

Update for 1768 and 1770, plus a couple other fixes:

Spoiler: v1.003 results

	Log Rating			Record
Player	Town	Scum	Combined	Town	Scum	Combined
GuyInFreezer	1.64		1.64	12-3	1-1	13-4
Thor665	0.46	-0.40	0.86	12-13	6-3	18-16
Malakittens	0.85		0.85	17-10	3-3	20-13
fferyllt	0.76		0.76	12-5	3-1	15-6
T S O	0.71		0.71	7-5	2-2	9-7
Xayzeck	0.58		0.58	7-4	1-4	8-8
Micc	0.57		0.57	5-4	2-1	7-5
singersigner	0.55		0.55	6-5	1-0	7-5
Huntress	0.55		0.55	7-5	1-3	8-8
Drixx	0.48		0.48	11-10	3-2	14-12
JaeReed	0.39		0.39	5-4	0-0	5-4
RadiantCowbells	0.34		0.34	14-15	4-1	18-16
Hopkirk	0.33		0.33	5-3	2-0	7-3
Bulbazak	0.28		0.28	5-4	2-2	7-6
Loopdan	0.26		0.26	5-5	2-0	7-5
Guyett	0.26		0.26	7-7	2-2	9-9
Raskolnikov	0.23		0.23	5-6	1-2	6-8
tojam2	0.17		0.17	3-5	0-1	3-6
theslimer3	0.09		0.09	4-4	0-0	4-4
jmo16mla	0.08		0.08	5-5	0-0	5-5
hayatoBL	0.03		0.03	5-6	3-1	8-7
TheIrishPope	0.00		0.00	4-5	1-3	5-8
innocentvillager	-0.02		-0.02	5-7	4-1	9-8
Wisdom		0.04	-0.04	3-4	6-4	9-8
copper223	-0.10		-0.10	3-5	2-0	5-5
goodmorning	-0.20	-0.09	-0.11	14-23	7-5	21-28
Nachomamma8	0.70	0.83	-0.13	17-12	5-12	22-24
House	-0.18		-0.18	5-7	2-0	7-7
Nobody Special	-0.19		-0.19	6-8	1-2	7-10
JasonWazza	-0.19		-0.19	4-5	3-3	7-8
RachMarie	-0.19		-0.19	11-18	2-2	13-20
Dierfire	-0.21		-0.21	6-10	2-2	8-12
Sakura Hana	-0.22		-0.22	5-6	3-3	8-9
Accountant	-0.23		-0.23	6-11	2-1	8-12
Not_Mafia	-0.30		-0.30	6-10	3-4	9-14
PhantomCobalt	-0.32		-0.32	5-8	1-1	6-9
BlueBloodedToffee	-0.13	0.21	-0.34	10-17	6-5	16-22
Ms Marangal	-0.41		-0.41	3-6	1-4	4-10
NicCage	-0.45		-0.45	2-6	0-0	2-6
Cabd	-0.46		-0.46	6-8	0-0	6-8
notscience	-0.29	0.19	-0.48	9-16	5-4	14-20
Alexcellent	-0.49		-0.49	3-5	0-1	3-6
enomis	-0.51		-0.51	2-6	0-0	2-6
Flubbernugget	-0.70		-0.70	3-8	0-0	3-8
MarioManiac4	-0.87		-0.87	2-8	1-2	3-10
Jake from State Farm	-0.91		-0.91	2-10	1-3	3-13
Titus	-1.00		-1.00	3-12	0-2	3-14

intercept: -0.403 (40.1% town win odds given entirely "other" players)
combined error: 248.87 (average 0.624)

Spoiler: updates and corrections list

Added newbies 1768 and 1770
RC added to newbie 1711
Added tojam2 to tracking list (has now reached 8 town games)

Interesting changes and other notes:
Nothing in particular from this update tbh

I should play more newbies...

[mhsmith0]

I should play more newbies...

FWIW you're actually close to getting on the ratings list; my #s have you as 5-1 town, 1-2 scum. So two more town games gets you to 8 and a rating.

[Cerberus v666]

I wish I liked newbie games just so I could be better than Drixx.

[Realeo]

I was compiling my new algorithm until something struck me....

Why are we using logarithmic loss?

[Realeo]

Omg. I'm stupid.

I have been too harsh at myself by saying "Realeo, your algorithm is worse than coin toss" until I realize that coin toss is discrete algorithm while everything else is a continuous algorithm. I can't compare discrete to continuous.

[Realeo]

A coin toss is discrete so I make my program spits discrete prediction. The new algorithm has an accuracy of 51.7%

over cross-validation

After I have actual free time, I will deploy the leaderboard of the new alogirthm.

Be assured that I still compile Open. However, to get insight why it is so hard to compile, I challenge you to compile open 626.

You will sympathize for me after opening Open 626

[rb]

so have we discerned who is the most best?

[Realeo]

Not yet. We are now trying to find which one is the best to find which one is the best.

[Realeo]

So I finally find myself a free time, but I have a question.

Say that you have a boolean equation in Excel. How do you apply those equation to the entire row? It's getting annoying if I have to click and drag the whole thing--especially the fact that I may have to do that repeatedly.

[mhsmith0]

Can you give a more detailed example? Not sure what you're asking.

[Realeo]

I have an =NOT(AND()) formula that I want to apply to the entire column. I usually put the formula at one cell and drag it down. It becomes cubbersome.

[Realeo]

I have only finished moving the 2014 data

It employs different algorithm--glicko. New explanation comes soon.

[mhsmith0]

If you double click the bottom right if a cell, it may copy all the way down. Or could copy cell, then control shift end, then shift+left until you're back to where you want to be, which usually takes just a couple seconds for a copy.

Or you could just do =and(whole range)? Still not quite sure what you're trying to do, sorry.

[Realeo]

I keep antagonizing myself for not sparing time to run the script to make the excel version.

So I spared some time and realize I already did. =(

[Realeo]

It's here! Updated Elo is here!

We use Glicko method to calculate your elo.

People kept protesting that

[X] Mafia is a small sample game
[X] We discard Theme game (and I fail to compile Open) diminishing a huge room of sample on our calculation.
[X] Every setup is different.
[X] It relies on other player as well
[X] Good player can be fucked up by bad player.

With so much noise, we cannot provide accurate elo.

Therefore, instead of giving an elo, we provide you a range of elo. If your elo is 1525

+

150, this means that your elo is in range of (1375,1675)

Your range goes smaller as you play game and goes higher as you took hiatus.

The elo is calculated every quarter.

More detail tommorow after well-deserved sleep.

[Psyche]

dataset??