Donk's ELO Point Spread System

[CamNewtonOutfit]

Hey donk! one question are you sure that ELO predicts that spread or is that just a spread that is the opening line? The reason i ask is that I have seen cases where 538 actually checks the box where it predicted the winner but the winning team is not the team that covered the spread (so they lost the predicted but checked the box anyway).

Last year I closely followed point spreads generated by FiveThirtyEight and decided to continue this year. For those of you who don't know, FiveThirtyEight is a website that does countless numbers of projections including NFL. They have an Elo point spread system where they generate a point spread for every NFL game based on team skill level, previous games, and locations of each game.

Since I like to crunch numbers when I get bored, I decided to compare the Elo point spreads with the closing odds from Vegas. As a result, I created 3 types of wagers based on the system. I was not planning on posting the results of this system anytime soon, however, after 3 weeks of NFL I actually have some interesting results. Here are the three types of wagers...

FOX: name derived from FiveThirtyEight's logo; this wager is when Elo has the same projected winner as Vegas, but Elo has a larger spread
PIG: first three-letter animal I could think of; this wager is when Elo has the same projected winner as Vegas, but has a smaller spread than Vegas.
DOG: named because this wager is technically the underdog; this wager is when Elo and Vegas have different projected winners based on the spread.

If this is confusing right now, I promise there will be much more clarity once the results are posted. Again, this system was only created because I got bored one night and simply started crunching numbers. I have created a few "systems" in the past, including a MLB first inning system being very successful. However, I have also had some systems completely flop.

As of now, this system is being entirely used for tracking but feel free to tail if you would like. Depending on how the system progresses throughout the season, I may lay down real wagers. Questions, comments, and concerns are more than welcome. Without further ado...

[Notorious_Donk]

Hey donk! one question are you sure that ELO predicts that spread or is that just a spread that is the opening line? The reason i ask is that I have seen cases where 538 actually checks the box where it predicted the winner but the winning team is not the team that covered the spread (so they lost the predicted but checked the box anyway).

Whats up Cam

These spreads are numbers stated explicitly by 538 based on their own system. They take into account past games, how teams performed against certain opponents, where the game was played, etc. They do not look at anything outside of the field, whether that be media, news releases, injuries, or other external factors. The spread put out by 538 on Tuesday morning will be the same exact spread five days later on Sunday morning before kickoff. This is mainly why they consider it a "benchmark"

[CamNewtonOutfit]

Thanks for the info! is that your real photo?

[Notorious_Donk]

Thanks for the info! is that your real photo?

That would be Lil Wayne, Green Bay's #1 fan

[CamNewtonOutfit]

Cool! too bad about the panthers this year had a real meltdown in the secondary.

[CamNewtonOutfit]

Honestly its odd that the panthers are middle of the pack in the ELO rankings while the NFL power rankings has them all the way at the bottom, which one you think is correct? i honestly tend to agree with NFL rank.

[Notorious_Donk]

Honestly its odd that the panthers are middle of the pack in the ELO rankings while the NFL power rankings has them all the way at the bottom, which one you think is correct? i honestly tend to agree with NFL rank.

The ELO ranking may have something to do with the team they were last year. Although they havent been great this year, there are plenty of games left and enough talent on that team

[Time is Money]

Honestly its odd that the panthers are middle of the pack in the ELO rankings while the NFL power rankings has them all the way at the bottom, which one you think is correct? i honestly tend to agree with NFL rank.

Power rankings mean nothing dude, just a bunch of knee jerk reactions.

I mean they're laying 2.5 to a Cards team that should make the offs.

Books haven't wrote them off just yet, they are still getting respect.

[Notorious_Donk]

NFL Week 7 Results

FOX
Bills -2.5 LOSS
Chiefs -6.5 LOSS
Falcons -5 LOSS

PIG
Bears +7.5 LOSS
Browns +11.5 LOSS
Eagles +3 WIN
Seahawks +1.5 WIN
Texans +8.5 LOSS

DOG
Colts +4 WIN
Rams +3 LOSS
Raiders +2 WIN
Lions +1 WIN

NFL Week 7FOX = 0-3
PIG = 2-3
DOG = 3-1
------------------
TOTAL = 5-7 (41.67%)

YTD Results
FOX = 17-16
PIG = 15-19
DOG = 9-3
------------------
TOTAL = 41-38 (51.90%)

FOX/DOG TOTAL = 26-19 (57.78%)

[BigdaddyQH]

52% is a loser. This is interesting, but really not worth the time nor the wagering.

[Notorious_Donk]

52% is a loser. This is interesting, but really not worth the time nor the wagering.

It's interesting enough to take up 20 minutes of my day. As of now the only thing that would be worth wagering is Dog and/or Fox plays. But either way i'm not putting any money on these plays. I'm sticking to my own plays for at least a couple more weeks

[Notorious_Donk]

NFL Week 8 Elo Spreads

Titans -2.5
Bengals -0.5
Panthers pk
Texans -2.5
Chiefs -3.5
Patriots -0.5
Jets -4.5
Buccaneers -1.5
Seahawks -6
Broncos -9
Falcons -1.5
Cowboys -2.5
Vikings -7.5

[gojetsgomoxies]

great stuff.... i am working on something similar. tracking kambour.net picks which are power rating based.

i think what you are tracking - and i'm working on - is whether the line under or over adjusts for recent info. not sure if there is constant bias too i.e. do point spread systems pick alot of big underdogs?

i think your systems generate enough picks that you don't really need to take "not adjusted for injury" games out of the picks....... i would think if you are going to take the games out that you may have to take NE games out for the next few games........ on second thought, maybe it is smart to take those games out as they don't capture what you are trying to measure.

one question: is there anything to suggest that nate silver's site (forgot exact name) does a more comprehensive, different and/or better job than sagarin, teamrankings.com, statfox, massey, howell etc. etc.?........... does anyone do quant power ratings and then qualitatively adjust at the end?

[gojetsgomoxies]

52% is a loser. This is interesting, but really not worth the time nor the wagering.

he's tracking..................... if you want a 55%+ system/wagering then it's alot of work to come up with it

plus 2 of 3 categories are 58%

[gojetsgomoxies]

surprised you have no pushes. or did i miss that?

i do think you need to get a bit more objective on the lines you use.... open? close? average of them?

[Notorious_Donk]

great stuff.... i am working on something similar. tracking kambour.net picks which are power rating based.

i think what you are tracking - and i'm working on - is whether the line under or over adjusts for recent info. not sure if there is constant bias too i.e. do point spread systems pick alot of big underdogs?

i think your systems generate enough picks that you don't really need to take "not adjusted for injury" games out of the picks....... i would think if you are going to take the games out that you may have to take NE games out for the next few games........ on second thought, maybe it is smart to take those games out as they don't capture what you are trying to measure.

one question: is there anything to suggest that nate silver's site (forgot exact name) does a more comprehensive, different and/or better job than sagarin, teamrankings.com, statfox, massey, howell etc. etc.?........... does anyone do quant power ratings and then qualitatively adjust at the end?

I'm not sure if its necessary to remove NE games because they still played well with the two backups. If they would have went 0-4 then it would have been a whole different story. However, I did remove the pick last week due to injury because Elo predicted Steelers by 1 point against Patriots even with Big Ben out. The elo spread didnt take into account Big Ben being injured, so it generated the spread thinking he was going to be under center.

As of right now, there is no solid evidence to prove that this site does a better job than any site out there. This is just a site I stumbled upon in the past and thought it was a good benchmark.

[Notorious_Donk]

surprised you have no pushes. or did i miss that?

i do think you need to get a bit more objective on the lines you use.... open? close? average of them?

I have had a few pushes but not that many. If you look back a few weeks you can see a couple games that resulted in a push.

Typically I have been using the closing line or a line an hour or two before kickoff

[gojetsgomoxies]

I have had a few pushes but not that many. If you look back a few weeks you can see a couple games that resulted in a push.

Typically I have been using the closing line or a line an hour or two before kickoff

donk, thx for both responses........ thought you were using local bookies line for your spread. if so, it would be hard to go back and check other stuff as you might not have record of it all.

i figured you went with nate silver's picks just for convenience/ease of use etc... same reason i went with kambour.net. you might want to check out his site........... was basically curious also if nate does it obviously DIFFERENTLY than others and it seems like probably.

i see sagarin - i would say the king of power ratings - now has forecasts for each game.

was confused at first by your characterizations but as you said once you dive into it it's pretty straight forward....

i like the terminology "favorite" (spread too small) and "underdog" (spread too big) and "wrong team"

i get the feeling that power rating based big is quite biased to favorites - maybe college though more

[gojetsgomoxies]

actually it looks like you have more underdogs than favorites....... the underdog value picks and the wrong team favored picks > favorite picks.

[gojetsgomoxies]

i did another system where i faded point spread moves but i had too many situations where the line probably moved because of QB injury/suspension info which was definitely NOT what my system was trying to capture.

[Notorious_Donk]

donk, thx for both responses........ thought you were using local bookies line for your spread. if so, it would be hard to go back and check other stuff as you might not have record of it all.

i figured you went with nate silver's picks just for convenience/ease of use etc... same reason i went with kambour.net. you might want to check out his site........... was basically curious also if nate does it obviously DIFFERENTLY than others and it seems like probably.

i see sagarin - i would say the king of power ratings - now has forecasts for each game.

was confused at first by your characterizations but as you said once you dive into it it's pretty straight forward....

i like the terminology "favorite" (spread too small) and "underdog" (spread too big) and "wrong team"

i get the feeling that power rating based big is quite biased to favorites - maybe college though more

I do try to use my local bookie for the lines, but if I miss the closing lines then I use vegasinsider

[Notorious_Donk]

I was out of town for a while so couldn't make any Elo picks. I will put in some time this morning to catch up and then post Week 10 plays

[Notorious_Donk]

NFL Week 8 Results

FOX
Texans -2.5 WIN
Jets -2.5 WIN
Bucs pk LOSS
Seahawks -1 LOSS
Broncos -3.5 WIN
Vikings -5.5 LOSS

PIG
Redskins +3 WIN
Cardinals +2.5 LOSS
Packers +3 WIN
Eagles +5 LOSS

DOG
none

YTD Results
FOX = 20-19
PIG = 17-21
DOG = 9-3
---------------------
TOTAL = 46-43
FOX/DOG TOTAL = 29-22 (56.86%)



*PS: I scratched the Bills-Patriots game because Brady did not play in the first matchup this season.

[Notorious_Donk]

NFL Week 9 Results

FOX
Cowboys -7 WIN
Chiefs -7 LOSS
Chargers -4 WIN
Broncos pk LOSS

PIG
Bucs +4.5 LOSS
Jets +4 PUSH
Ravens +3 WIN
Rams +3 PUSH
49ers +5.5 LOSS

DOG
Eagles +3 LOSS

Week 9 = 3-5-2

YTD Results
FOX = 22-21
PIG = 18-23
DOG = 9-4
--------------------
TOTAL = 49-48
FOX/DOG TOTAL = 31-25 (55.36%)

[Notorious_Donk]

NFL Week 10 Elo Spreads

Ravens -10
Eagles -1
Bucs -2.5
Broncos -2.5
Packers -4.5
Texans -5.5
Chiefs -0.5
Jets -4
Redskins -1
Chargers -3.5
Steelers -0.5
Cardinals -12
Patriots -4.5
Giants -2

[Notorious_Donk]

It was a Trump sized HUUUUUGE week for the system and we finished it off with a SNF winner and a MNF winner. After a very modest 9 weeks, the system finally broke through and provided some real value. I am still not betting these plays, but after this week I may take a closer look. Here are the results...

NFL Week 10 Results

FOX
Ravens -7.5 WIN
Packers -3 LOSS

PIG
Vikings +2.5 LOSS
Dolphins +4.5 WIN
Cowboys +3 WIN
49ers +14 WIN
Seahawks +7.5 WIN

DOG
Bucs +2.5 WIN
Broncos +3 WIN
Texans +3 WIN
Chiefs +3 WIN
Jets +1 LOSS
Giants +1 WIN

FOX = 1-1
PIG = 4-1
DOG = 5-1
--------------------
TOTAL = 10-3
YTD TOTAL = 59-51
FOX/DOG TOTAL = 37-27 (57.81%)

[Notorious_Donk]

NFL YEAR TO DATE RESULTS

Week 1 = 5-2
Week 2 = 7-3
Week 3 = 4-0
Week 4 = 4-4
Week 5 = 1-4
Week 6 = 2-2
Week 7 = 3-4
Week 8 = 3-3
Week 9 = 2-3
Week 10 = 6-2
--------------------
TOTAL = 37-27 (57.81%)

[Notorious_Donk]

NFL Week 11 Elo Spreads

Panthers -5.5
Vikings -1
Cowboys -8
Bengals -1.5
Giants -7.5
Lions -11
Steelers -7.5
Chiefs -11.5
Colts -5.5
Rams -1.5
Patriots -10.5
Seahawks -7
Redskins -4
Texans pk

[Notorious_Donk]

NFL Week 11 TNF

Panthers are currently 3.5 pt favorites and will be a FOX play as long as the line stays under 5. Watch the line movement this afternoon

[CamNewtonOutfit]

hey donk bad news for ELO, i did some research and noticed that for the NBA 2015-2016 season ELO spread was a loser, meaning lets say you took a bet where ELO said spread was below vegas spread so you bet against the spread that would be a consisent loser for the first 20% of the season ( have not checked the rest of the season just the first 20% but so far its down 10%. Graph attached.

[CamNewtonOutfit]

Have you tried to analyze nba results?

[Notorious_Donk]

hey donk bad news for ELO, i did some research and noticed that for the NBA 2015-2016 season ELO spread was a loser, meaning lets say you took a bet where ELO said spread was below vegas spread so you bet against the spread that would be a consisent loser for the first 20% of the season ( have not checked the rest of the season just the first 20% but so far its down 10%. Graph attached.

Completely different sports tho. Way too much inconsistency in NBA from day to day.

I would expect a system like Elo to predict football at a much higher rate since there is a smaller testing population and much less variance

[Notorious_Donk]

NFL Week 11 Sunday

FOX
Cowboys -7
Lions -5.5
Chiefs -7
Colts -3
Rams pk
Redskins -3

DOG


PIG
Cardinals +2
Bears +6.5

No dog plays today but there are 6 fox plays. I will look at the lines closer to noon kickoff and reevaluate

[Notorious_Donk]

NFL Week 11 MNF

DOG: Texans +6

[Notorious_Donk]

NFL Week 11 Results

FOX
Panthers -3.5 LOSS
Cowboys -7 WIN
Lions -5.5 WIN
Chiefs -7 LOSS
Colts -3 WIN
Redskins -3 WIN

DOG
Rams +1 LOSS
Texans +6.5 LOSS

PIG
Cardinals +2 LOSS
Bills +2.5 WIN
Browns +9 LOSS
49ers +12 LOSS

Week 11 Total (no pig plays) = 4-4

YTD Total = 41-31 (56.94%)