NHL Handicapping System: PythagLog

This is a system I've developed over the last 2 years by combining statistical models and data analysis from both Bill James's Pythagorean Expectation models, Log5 models and Alan Ryder's hockey analytics website.

In a nutshell, I use a pythagorean expectation model to determine an expected winning percentage for each team and then the Log5 model to determine win probabilities for each of the two teams.

I then compare the the expected winning percentage to actual winning percentage by using the real-time team record and then calculate a log5 winning percentage for each team.

These winning percentages can then be used to calculate value bets for either the favourite or the dog by using them in conjunction with the current odds.

Let's walk through one together:

*******************

Jan 4, 2016 DET @ NJD

First we determine the goals scored for and against each team specific to the home/road situation they are facing.

Detriot (away) -  40 GF, 45 GA, 16 away games

New Jersey (home) - 43 GF, 46 GA, 20 home games

We then plug these numbers into the pythagorean expectation model using 2.15 as the exponent (as per Alan Ryder's hockey analytics).

The Pythagorean Expectation Model

(GF^2.15)/(GF^2.15+GA^2.15)

So we plug these in and we get:

Detroit: 0.437 Expected Win % (aka 6.992 games)

New Jersey: 0.464 Expected Win % (aka 9.276 games)

We then use the log5 model to predict how these two teams would fare against one another.

Log5 Winning Expectation Model

(A-A*B)/(A+B-2*A*B)

A: Team 1 Winning %
B: Team 2 Winning %

So we do this for each team and get the following:

Detroit: 0.473 (47.3%)

New Jersey: 0.527 (52.7%)

You can take win percentage and look for angles and values against the current odds.

What I like to do next though is to perform a Log5 calculation using the teams actual records ( as opposed to their expected records). This lets me know if a team is over-performing or under-performing their winning expectation.

Detroit (away) - 8 away wins ( 0.500 Actual win %)

New Jersey (home) - 9 home wins (0.450 Actual win %)

As you can see, Detroit is over-performing their winning expectation while New Jersey is slightly under-performing theirs.

Next I perform a  Log5 calculation on these actual win percentages:

Detroit : 0.550 (55%)

New Jersey: 0.450 (45%)

Finally, I look at the current odds and determine if there are any value bets to be had either for the favourite or the dog.

Generally I want to see a team that has an actual win % that is higher than the expected win % number.

There are other angles also.

Last year my rules were:

1. Away favourite
2. Expected win % higher than 51%
3. Actual win % higher than expected win %
4. Positive team strength (more GF than GA)
5. Positive bet value expectancy

I did very nicely.

Hope this helps. If you have questions, suggestions, new angles or improvements to this please let me know!

I like it! Best of luck.

Will you be posting the plays on here?

The use of numbers attached to a formulated system is a great approach.

Appreciate the feedback! If you have any ideas to make this better via filtering, MM strategy, or otherwise please let me know.

I also like to make sure that the starting goaltender is the team's regular starter -- I've been burnt a few times by betting game with good numbers where the coach throws in the backup...Boston with Jonas Gustavsson in net comes to mind lol

Montreal the last little while is a perfect example -- they are simply not the same team without Price in the net.

I also have a pet theory that away favourites that have the stats I'm looking for are the way to go because if your GF/GA on the road are that strong it means that your skill outweighs any home ice advantage. This is just speculation on my part though.

Here are my stats for today:

Format: Team: Expected Win % / Actual Win %

NJD: 52.7% / 45.0%
DET: 47.3% / 55.0%
No bet, both teams have negative team strength

STL  58.1% / 68.4%
OTT 41.9% / 31.6%
No bet, both teams have negative team strength

COL 26.6% / 21.2%
LAK 73.4% / 78.8%  32.16% Value
Bet LAK on Moneyline

EDM 62.1% / 65.4%  21.31% Value
CAR 37.9% / 34.6%
Bet EDM on Moneyline

VAN 60.9% / 51.2%
ARI 39.1% / 48.8% 
No bet, both teams have negative team strength

@DanRules24: Ive started testing this in conjunction with your 2 road team parlay system to see if handicapping for the two strongest teams of the day nets more profit than simply the two best priced road teams.

Last night my system would have suggested parlaying FLA and CHI, both of whom won their games.

@Oilcountry99: Thanks for the feedback. I would agree that the filtering is hugely important!

Tormoid - Goaltending is huge, glad you include that in your assessment 

I'm rooting for ya Tormiod!

Using a flat bet MM strategy at the moment, but I'm looking at a few Tony Stoffo chase systems as well.

Here is my somewhat automated spreadsheet for making calculations.

Simply type in each team's GF, GA and games played in the home/away situation and it will calculate expected Win % and expected wins.

If you then go down the page and enter the number of actual games won for each team in the home/away situation it will calculate actual Log5 win % as well.

Finally, if you enter your preferred book's odds in decimal format you can see bet values for both teams.

Enjoy!

https://www.mediafire.com/download/lsrz89qo3bod0of/PythagLog+NHL.xlsx

Ahhhh, Remember ,there is always intangibles. This may work early in the season but as it wears on, burn out is huge. I've been officiating hockey for 30 years in all levels except NHL. The players are burnt ,the coaches are fried, and worst of all the refs are smoked. You should time frame your system on a monthly basis to see if it correlates later in the season. Just saying. Good luck.

Great point. I actually find that its a much stronger model later in the season as it tends to work better with more GF/GA data. Its prime point seems to be 20 games of home/road data (aka mid point of the season).

Haven't tested by month or by way points during the season though -- thanks for the idea. I'll have to take a look at that.

Roger that. Keep me posted.

Stats for Tuesday January 5th:


BOS: 30.3% / 25.7%
WAS: 69.7% / 74.3% 29.26% Value

BUF: 29.9% / 31.4%
FLA: 70.1% / 68.6%
No bet, FLA actual win % lower than exp win %

NYR: 60.6% / 63.6% 14.14% Value
DAL: 39.4% / 36.4%

PIT: 62.0% / 57.9%
CHI: 38.0% / 42.1%
No bet, both teams have negative strength

CBJ: 49.1% / 56.0%
MIN: 50.9% / 44.0%
No bet, both teams have negative strength


PHI: 36.5% / 44.7%
MTL: 63.5% / 55.3%
No bet, MTL actual win % lower than exp win %

NAS: 81.6% / 76.3% 29.75% Value
WPG: 18.4% / 23.7%

CGY: 46.0% / 65.6%
TBL: 54.0% / 34.4%
No bet, both teams have negative strength

I'm on a 2 team parlay -- NAS / WAS.

Good luck tonight everyone!

Not many good games for tonight, but here we go:

MTL: 57% / 50%
NJD: 43% / 50%

CHI: 77.4% / 71.4%
PIT: 22.6% / 28.6%

COL: 45.4% / 34.4%
STL: 54.6% / 65.6%

VAN: 54.8% / 48.8%
CAR: 45.2% / 51.2%

ANA: 60.7% / 60.4%
TOR: 39.3% / 39.6%

I'll be on a 2 team parlay tonight of ANA / STL.

Good luck everyone

Another lacklustre performance. Last night.

Last year following my rules with this spreadsheet I was around 77% win rate and up about +120 units. This year over 50 bets or so I've been 44%....just getting hammered.

Until I find a way to improve this system I'm going to stop posting picks...as so far this year just doing the opposite of what the system suggests would have yielded better results.

I'm wondering if only using the last 10 games of data would catch which teams are trending?

I may have to devise an expected goals model.

In the meantime I think I'll be following danrules24 with his road team parlay chaser.

Okay I've made some changes to the algorithm. Let's see how it goes!

Today's Game Analysis
**************************
NYR: 2.858
PHI: 2.473
NYR 56.4% // TOTAL 5.331 GOALS
7.16% Edge. M\L Bet
******************************

NJD: 2.688
ARI: 2.809
ARI 52.0% // TOTAL 5.497 GOALS
-8.32% Edge. No Bet.
******************************

OTT: 2.244
LAK: 3.052
LAK 63.4% // TOTAL 5.297 GOALS
-5.8% Edge. No Bet.
******************************

TOR: 2.557
BOS: 2.731
BOS 52.9% // TOTAL 5.288 GOALS
-16.8% Edge. No Bet.
******************************

WAS: 3.335
BUF: 1.539
WAS 79.1% // TOTAL 4.894 GOALS
25.6% Edge. M/L Bet.
******************************

COL: 3.506
CBJ: 2.484
COL 65.8% // TOTAL 5.990 GOALS
27.2% Edge. M/L Bet
******************************

MTL: 2.796
STL: 2.328
MTL 58.0% // TOTAL 5.124 GOALS
23.54% Edge. M\L Bet.
******************************

MIN: 2.113
NAS: 2.762
NAS 61.3% // TOTAL 4.875 GOALS
11.13% Edge. M/L Bet.
******************************

CGY: 2.536
EDM: 3.385
EDM 63.3% // TOTAL 5.921 GOALS
24.13% Edge. M/L Bet.
******************************

DAL: 3.507
SJS: 2.636
DAL 63.4% // TOTAL 6.143 GOALS
38.21% Edge. M/L Bet.
******************************

I'll bet paper trading these picks for a while to see how the newly tweaked algorithm does.

Saturday Results:
*******************
NYR Win
BUF Win
CBJ Win
STL Win
NAS Win
EDM Win
SJS  Win

******************************
3-4

Sunday's Game Analysis
***************************

CAR: 2.666
PIT: 2.506
PIT: 50.1% // TOTAL GOALS 5.172
UNDER 5.5: 58.6%
No Bets.
********************************

VAN: 2.355
NYI: 3.269
NYI: 72.7% // TOTAL GOALS 5.624
UNDER 5.5: 50.8%
M/L NYI 14.7% Edge
********************************

FLA: 2.509
TBL: 2.179
FLA: 56.2% // TOTAL GOALS 4.688
UNDER 5.5 67.1%
No Bets.
********************************

NYR: 2.031
WAS: 3.678
WAS: 81.9% // TOTAL GOALS 5.709
OVER 5.5: 50.7%
M/L WAS 32.1% Edge
********************************

MTL: 2.008
CHI: 3.125
CHI: 76.6% // TOTAL GOALS 5.132
UNDER 5.5: 59.3%
M/L CHI 19.8% Edge
Under 5.5 3.1% Edge
********************************

PHI: 2.401
DET: 2.746
DET: 63.9% // TOTAL GOALS 5.146
UNDER 5.5: 59.0%
No Bets.
********************************

LAK: 2.694
ANA: 1.908
LAK: 71.3% // TOTAL GOALS 4.602
UNDER 5.5: 68.5%
M/L LAK 42.6% Edge
Under 5.5 3.4% Edge

Sunday's Game Results
***************************

VAN WIN

WAS WIN 

CHI WIN   
OVER WIN

LAK WIN    
UNDER WIN

Sunday Total Results:
4-2