Project ELO (Dataset released)

[Realeo]

I cannot release the elo algorithm yet because it still relying on the not-so-cleaned data. (ie. I haven't create the parser of cleaned data)

[Realeo]

The data has been revised. The data for Normal was inaccurate and has been rectified to reflect the actual representation.

[Realeo]

Test result is coming!

The lower the score, the better

ELO: 0.754232774676
mhsmith factor: 0.697544079138
Coin toss : 0.69314718056

Gah! Both multivariate-and-seperated with ELO-but-combined is both inferior to

fair coin toss

With that...Research and Developemnt Continues.

[Realeo]

I think untill we find an algorithm that is better than coin toss, we shouldn't update either our leaderboard until we find an algorithm that is better than a coin toss.

I will be angry if the optimal algorithm is a neural network =/

[Realeo]

I think untill we find an algorithm that is better than coin toss, we shouldn't update either our leaderboard

It doesn't mean we shouldn't update our algorithm.

[Realeo]

It's funny that on the first run, I accidentally used the wrong K factor and actually performed better than a coin toss.

[Realeo]

Apparently, mhsmith, I misread the way your multivariate algorithm works, thus I re-coded again and re-test it again.

Your error is now +0.677034669602, which is

better

than coin tos.

Congratulation! Currently, I'm programing the ELO which town/scum is seperated. It's done. Preparing to train the K factor.

[mhsmith0]

I'm kinda curious why I'm st 0.625 and you're at 0.677 from my #s. I'll probably post a google doc w my calcs and see where that goes.

[mhsmith0]

I'd also suggest were probably never gonna get WAY better than coin toss, since there are VERY few players who are way better than rand as both alignments (ffery stands out in newbie games as town and scum, no one else really does as both).

[Realeo]

I'm kinda curious why I'm st 0.625 and you're at 0.677 from my #s. I'll probably post a google doc w my calcs and see where that goes.

We're using Normals data, right?

[mhsmith0]

Oh no, I'm just using newbies data entirely (I'm ignoring normal games due to the more substantial variation in balance game to game, as well as the occasional presence of multiball and/or sk in normals/opens). If that's the difference then that makes sense.

[fferyllt]

I'd also suggest were probably never gonna get WAY better than coin toss, since there are VERY few players who are way better than rand as both alignments (ffery stands out in newbie games as town and scum, no one else really does as both).

I am never going to play another newbie game out of fear of ruining this, whatever it is.

[mhsmith0]

Kinda hard to argue with a 12-5 town record and 3-1 scum record. not sure what your pre-matrix6 stats would look like, but it's a very impressive couple of year run.

Edit: actually I think there were matrix6 games before 1370, but no one tracks who was what faction and I don't care NEARLY enough to manually do all that work lol

[fferyllt]

Offhand I think all my pre-matrix 6 games except for the first one were town games, and I *think* they were all wins. Weren't too many of them.

The last several months of 2015 I was playing terribly and getting stomped in many of my non-Newbie games. I lost my mojo.

[mhsmith0]

[mhsmith0]

FWIW, I think that most people, to the extent that they succeed notably as a player (obviously there are plenty of people who don't much succeed as any alignment), tend to be mainly successful in one alignment or the other. Nacho is an extremely obvious example of this (I've played with him a LOT on MU, and he's amazing as town, but struggles at times as a wolf), and there are a bunch of EMers who've come on-site who have done great as scum but struggle as town.

[fferyllt]

I think most people, including me, think I'm much better at town than at scum. If I had more scum Newbie games that would be more obvious in the numbers.

[mhsmith0]

Yeah it's actually pretty hard to accurately ping someone's scum abilities in newbie games unless they play a LOT given the sheer rand of it.

Then again I'm 1-0 as scum in newbie so clearly I'm amazing at scum (unless you read basically any of my scum games lol)

It's actually kind of funny: people who see me as town tinfoil about me being a good wolf, except I'm consistently in the range of mediocre to hilariously bad as wolf, mainly because I'm just not a good or natural liar, and it often comes out to the point where people just mark me as obvscum without all that much effort.

[LicketyQuickety]

Yeah it's actually pretty hard to accurately ping someone's scum abilities in newbie games unless they play a LOT given the sheer rand of it.

Then again I'm 1-0 as scum in newbie so clearly I'm amazing at scum (unless you read basically any of my scum games lol)

It's actually kind of funny: people who see me as town tinfoil about me being a good wolf, except I'm consistently in the range of mediocre to hilariously bad as wolf, mainly because I'm just not a good or natural liar, and it often comes out to the point where people just mark me as obvscum without all that much effort.

I have no problem lying as Scum. I have difficulty in making myself look Town as Scum. My Town game is honestly so Fing townie that it really puts a hamper on my Scum game *shrug* Kinda funny considering I used to get lynch a ton as Town.

[Realeo]

Ahh! I hate tabulating Open!

[LicketyQuickety]

Ahh! I hate tabulating Open!

You got this homie.

[mhsmith0]

Coulda sworn I'd posted this already, but must have gotten closed before I hit submit. Update for 1766 and 1769, plus a couple other fixes:

Spoiler: v1.002 results

	Log Rating			Record
Player	Town	Scum	Combined	Town	Scum	Combined
GuyInFreezer	1.62		1.62	12-3	1-1	13-4
Thor665	0.44	-0.41	0.85	12-13	6-3	18-16
Malakittens	0.84		0.84	17-10	3-3	20-13
fferyllt	0.76		0.76	12-5	3-1	15-6
T S O	0.70		0.70	7-5	2-2	9-7
Micc	0.58		0.58	5-4	2-1	7-5
Xayzeck	0.58		0.58	7-4	1-4	8-8
singersigner	0.56		0.56	6-5	1-0	7-5
Huntress	0.54		0.54	7-5	1-3	8-8
JaeReed	0.53		0.53	5-3	0-0	5-3
Drixx	0.49		0.49	11-10	3-2	14-12
Hopkirk	0.31		0.31	5-3	2-0	7-3
Bulbazak	0.28		0.28	5-4	2-2	7-6
Loopdan	0.27		0.27	5-5	2-0	7-5
Guyett	0.25		0.25	7-7	2-2	9-9
Raskolnikov	0.25		0.25	5-6	1-2	6-8
RadiantCowbells	0.15		0.15	12-15	4-1	16-16
theslimer3	0.09		0.09	4-4	0-0	4-4
jmo16mla	0.07		0.07	5-5	0-0	5-5
hayatoBL	0.05		0.05	5-6	3-1	8-7
TheIrishPope	0.00		0.00	4-5	1-3	5-8
innocentvillager	-0.04		-0.04	5-7	4-1	9-8
Wisdom		0.05	-0.05	3-4	6-4	9-8
copper223	-0.10		-0.10	3-5	2-0	5-5
goodmorning	-0.21	-0.08	-0.12	14-23	7-5	21-28
Nachomamma8	0.70	0.84	-0.14	17-12	5-12	22-24
House	-0.16		-0.16	5-7	2-0	7-7
JasonWazza	-0.18		-0.18	4-5	3-3	7-8
Nobody Special	-0.19		-0.19	6-8	1-2	7-10
RachMarie	-0.20		-0.20	11-18	2-2	13-20
Dierfire	-0.21		-0.21	6-10	2-2	8-12
Accountant	-0.22		-0.22	6-11	2-1	8-12
Sakura Hana	-0.23		-0.23	5-6	3-3	8-9
PhantomCobalt	-0.28		-0.28	5-8	1-1	6-9
Not_Mafia	-0.30		-0.30	6-10	3-4	9-14
BlueBloodedToffee	-0.14	0.22	-0.37	10-17	6-5	16-22
Ms Marangal	-0.42		-0.42	3-6	1-4	4-10
NicCage	-0.45		-0.45	2-6	0-0	2-6
Cabd	-0.47		-0.47	6-8	0-0	6-8
Alexcellent	-0.47		-0.47	3-5	0-1	3-6
notscience	-0.30	0.18	-0.47	9-16	5-4	14-20
enomis	-0.51		-0.51	2-6	0-0	2-6
Flubbernugget	-0.71		-0.71	3-8	0-0	3-8
Titus	-0.84		-0.84	3-11	0-2	3-13
MarioManiac4	-0.85		-0.85	2-8	1-2	3-10
Jake from State Farm	-0.90		-0.90	2-10	1-3	3-13

intercept: -0.393 (40.3% town win odds given entirely "other" players)
combined error: 248.03 (average 0.625)

Spoiler: updates and corrections list

Added newbies 1766 and 1769
There was an issue where 1467 and 1710 were sort of being counted, those are now properly excluded entirely
Loopdan hadn't been included in 1767, he's back in

Interesting changes and other notes:
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
It's also interesting how Thor has a really high town rating given a slightly sub .500 record. That's probably a decent indicator of just how much towns tend to struggle in newbie games, where basically a .500 town record suggests that you're really good at being town in this setup.

[Gamma Emerald]

What happened to the one-year rule

[mhsmith0]

?

[Gamma Emerald]

"To be considered for ELO leaderboard, the player's last game must be finished 1 year ago." from the OP