ipredict - Football statistical predictions

[allen29]

Some new selections from today. 

Will provide an update this night or tomorrow on bankroll.

[allen29]

Bank: +5.09 (H: -16.67, A: +21.76) 

Yield: 2.4% (H: -20.6%, A: +16.6%)

Some selections from the last Round of Premier League and playoffs in lower leagues. Motivation should play some role here, especially for Sunderland, Norwich and Newcastle which is not yet included in my models, but odds look pretty good for them.

[allen29]

I enjoyed so much West Ham's win last night!

Here are some new selections, mostly from the Americas.

[allen29]

More selections for the next few days

 

[own3dripy]

Very interesting thread.

[allen29]

It was a really poor period, -8.58 pts down since my last post. and that makes an overall yield of -1.44% (H: -15%, A: 8.12%).

American (or Americas) leagues are causing a big loss, it is extremely hard to get away wins there. Even though I include home field advantage for each team, I think I underestimate the probability of home wins. The reason is because I use a lot of data from English and Scottish leagues where away wins are more frequent than MLS, Argentina, Brazil, Mexico. I don't know yer how will I solve this, but I will go easier on these leagues, and maybe start taking some DC's or Asian handicaps instead.

[allen29]

I've been tinkering a bit this weekend and added more historical data to the model. There was a bias for Away games, the selections were disproportionate (about 2/3 for Away games). Now, they should be more balanced. I will wait few more rounds for Argentina until I have maybe 20 historical games to have more reliable estimates of team quality.

I found a new data source, Onefootball ,that has the text for many more leagues and some will be played during summer. I'll add them maybe in 1 month, to busy at the moment.

I wanted to have a little dashboard for each game so I can visualize and get a sense of why there seems to be value on a certain side. Still experimenting how to format it and what things to include.

Long term strength is one component of my models. It is scaled 0-100. The average team has a Strength of 44 and it is only comparable within the same League. For example, Celtic and Rangers have a strength similar to Barcelona and Bayern, meaning that they have similar performances within their leagues.

Edited May 23, 2016 by allen29

[allen29]

Going a little more conservative with DC's and handicaps on MLS. So far, 53% of the games have ended with a Home Win there...this is huge home field advantage compared to European leagues: Premier League (41%), Championship (41%), League 1(43%), Scottish PL (41%), La Liga (48%), Bundesliga (44%). I wonder why is that....maybe long travelling distances cause away team to be more tired...maybe some jet lag? This article states that " MLS supporters--meaning fans, from casuals to ultras--are the best sports fans in America. The data supports it." which can only make me laugh. Some time ago, I included distance between the two stadiums in Europe to see if it has any impact on predictions and it didn't, but I'll try that for MLS as well.

[allen29]

New York City vs Orlando City, X2 @ 2.20 (Model: 1.82)

[allen29]

Next selections, will provide an update at the end of the week.

[allen29]

Edited June 1, 2016 by allen29

[real55555]

hey bro, seen that you put in a lot of work in this. My advice to you is only go for the leagues that you are familiar. Different leagues played differently, so instead of using the english and scottish tendencies, maybe you should work out a set of tendencies for these america teams. Better still, focus back on what you've been doing that have been yielding positive results. I know it is hard to resist when it is offseason now but just take it as a break for yourself

[allen29]

Overall: -5.69 (-2.29%), Home: -15.27 (-16.66%), Away: 6.58 (4.38%), Double Chance: 2.90 (41.38%)

I have to take a break from this anyway, I don't have time to work on it until the end of the month. I want to include as many leagues as I can find relevant data for, next season I will add more leagues from Europe (including 2nd and 3rd League from Germany and 2nd League from Spain and France) plus China and Chile and hopefully the web application will be up. I take league tendencies into account, I try to make the models learn the characteristics of each League. But I have some new ideas that I want to test out during summer. 

Until the next season begins, I'll still do Brazilian league, MLS and from next month maybe Scandinavian leagues, but probably more sporadic selections.

 

 

 

 

[allen29]

Some late selections, let's see how they play out

[lawalahmed]

I think your result out put is similar to Bairns selection that huge depend on away games, and i will also advice you to check result of both home and away separately also start your selection after four games to the season also to stop by March ending.....Just my thought

[allen29]

Hey, I'm back and I'll be posting new selections. I've been working really hard to add more leagues and some statistics that I had in mind and thought would improve the model. I think I did a decent job of including league characteristics so that the model doesn't under predict home wins in some leagues and can be generalized to different leagues.

Now I will cover about 30 leagues (not all at the same time), which I'm very excited about. Some rounds must pass in a league until I can start making predictions (as of today, I need every team to have at least 20 historical games and 6 games from the previous season to be included in the selections). @lawlahmed, I think you have point. I tested my models while excluding the first and last 15% of games from each league and accuracy and return improved, so I will stick to this rule for the future. I also want to do BTS and Over/Under bets but I still need to collect market odds to get an idea of the returns.

Here is a new visualization template that I've created. One thing it solves for me is that it tells me who are the most productive offensive players so I can look up faster if they are available or not for then next game. (XG + XA stands for expected goals + expected assists). Besides the probability distribution of market odds and my model, there are also 4 new ones. They are based on outcomes in similar games (teams with similar strength and form). The difference between "Similar" and "Similar2" is between what statistics I use to search for similar games. "Similar2.out" is the actual outcome in similar games, "Similar2.xout" is calculated based on win expectancy and not actual wins, and "Similar2.bmk" is the win expectancy from the bookmaker's odds. I don't use these very much as they are already variables included in the model.

Here is a game from tomorrow that looks tempting.

http://imgur.com/a/BesCu#OYotE4x

Edited August 7, 2016 by allen29
Image link was broken

[allen29]

Coritiba vs [Ponte Preta] (X2 @ 1.92 - Pinnacle). Juan and Alan Santos, 2 important offensive players from Coritiba are suspended. In addition, my models see some small value on the away team.

[Saarspborg] vs Valerenga (HW @ 2.31 - Pinnacle)

Atletico Nacional - [Once Caldas] 2 AH + 1 @ 2.55 Pinnacle

LA Galaxy - [NY Red Bulls] X2 @ 1.99 - Pinnacle

Stabaek vs [Haugesund] was at @ 3.40 - Marathon

 

[allen29]

Palmeiras - [Vitoria]  2 AH + 1.5 @ 1.70 - Pinnacle

 

[allen29]

New games for the weekend. Could be worth waiting a bit until more bookmakers are available.
 

Date	League	Match	Side	Bmk.odds	My.odds	Value	Bookmaker	Stake
8/11/2016	NOR	Viking FK vs Rosenborg	Home	4.44	3.52	0.26	Pinnacle	1
8/12/2016	CN	Beijing Guoan FC vs Chongqing Lifan	Draw or Away	2.59	2.12	0.22	Pinnacle	1
8/13/2016	BRA	Sport vs Flamengo	Home	2.74	2.34	0.17	Marathonbet	1
8/13/2016	MLS	Vancouver Whitecaps vs San Jose Earthquakes	Away	4.25	3.96	0.07	Marathonbet	0.5
8/14/2016	BRA	Atletico Paranaense vs Palmeiras	Away	2.98	2.67	0.12	Marathonbet	1
8/14/2016	BRA	Cruzeiro vs Coritiba FBC	Away	5.6	4.74	0.18	Marathonbet	0.5
8/14/2016	MLS	Houston Dynamo vs Toronto FC	Away	3.26	2.84	0.15	Marathonbet	1
8/14/2016	BRA	Santos FC vs Atletico MG	Home	2.4	2.18	0.10	Marathonbet	1

 

 

Edited August 10, 2016 by allen29

[allen29]

Few more after some price changes...

Date	League	Match	Side	Bmk.odds	My.odds	Value	Bookmaker	Stake
8/13/2016	BRA	AA Ponte Preta vs Figueirense FC	Draw or Away	2.26	2.01	0.12	Marathonbet	1
8/14/2016	BRA	Cruzeiro vs Coritiba FBC	Draw or Away	2.47	2.18	0.13	Pinnacle	1
8/13/2016	CN	Shanghai Shenhua vs Guangzhou Evergrande	Home	3.64	2.89	0.26	Marathonbet	1
8/14/2016	CN	Changchun Yatai vs Jiangsu Guoxin-Sainty FC	Home	3.85	2.72	0.42	Marathonbet	1
8/14/2016	CN	Guangzhou R&F FC vs Shandong Luneng	Away	3.56	2.67	0.33	Pinnacle	1
8/14/2016	MLS	Chicago Fire vs Orlando City SC	Draw or Away	1.94	1.81	0.07	Pinnacle	1

 

Edited August 12, 2016 by allen29

[allen29]

Hangzou Greentown - Henan Jianye (Away @ 3.78 - Pinnacle)

Caught this odd at 8/13/2016, 2:04. Now it has fallen back.

 

[allen29]

It was a great start! Bank chart

Since new model, 20 bets, 12 Won, 7 lost, 1 Voided. +13.53 pts, 71.23% yield. But there was for sure luck in these results.

 

 

[allen29]

Odds are from Tuesday night, there could be some changes.

Date	League	Match	Side	Bmk.odds	My.odds	Value	Bookmaker	Timestamp	Stake
8/20/2016	CN	Tianjin Teda vs Shandong Luneng	Away	2.64	2.15	0.23	Marathonbet	8/16/2016 22:22	1
8/21/2016	CN	Chongqing Lifan vs Guangzhou R&F FC	Away	3.22	2.8	0.15	Marathonbet	8/16/2016 22:22	1
8/21/2016	MLS	Colorado Rapids vs Orlando City SC	Draw or Away	2.88	2.18	0.32	Marathonbet	8/16/2016 22:22	1
8/21/2016	BRA	Coritiba FBC vs Santos FC	Home	3	2.35	0.28	Marathonbet	8/16/2016 22:22	1
8/21/2016	BRA	Figueirense FC vs Cruzeiro	Away	2.99	2.72	0.10	Marathonbet	8/16/2016 22:22	1
8/21/2016	CN	Jiangsu Guoxin-Sainty FC vs Beijing Guoan FC	Away or Draw	2.34	1.78	0.31	Marathonbet	8/16/2016 22:22	1
8/21/2016	ARG	Lanus vs Boca Juniors	Home	2.4	2	0.20	Marathonbet	8/16/2016 22:22	1
8/21/2016	ARG	Rosario Central vs Defensa y Justicia	Away	4.5	3.7	0.22	Marathonbet	8/16/2016 22:22	1
8/21/2016	MLS	Sporting Kansas City vs Vancouver Whitecaps	Home	1.91	1.82	0.05	Marathonbet	8/16/2016 22:22	1
8/21/2016	NOR	Stabaek vs Aalesunds FK	Away or Draw	2.2	1.94	0.13	Matchbook	8/16/2016 22:22	1
8/21/2016	ARG	Tigre vs Estudiantes La Plata	Away	3.36	2.85	0.18	Marathonbet	8/16/2016 22:22	1
8/26/2016	ARG	Belgrano de Cordoba vs Independiente	Away	3.56	2.67	0.33	Marathonbet	8/16/2016 22:22	1

 

 

Edited August 18, 2016 by allen29

[allen29]

1 more for tonight: 

Valerenga vs Bodo/Glimt , X2 @ 270, Pinnacle. Model has it at 2.04.

I checked the odds again and 10/12 dropped. Good sign for my selections I think, I always wondered if I should take a bet now, when the odds look very good, but only few bookmakers are available, or wait until more bookmakers open the event the odds may get even better. I think I will stick to betting very early if the odds are good enough as a few days later I might not catch them.

[allen29]

2 more for today or tomorrow depending on your time zone.

Date	League	Match	Side	Bmk.odds	My.odds	Value	Bookmaker	Timestamp	Stake
8/20/2016	COL	Boyaca Chico vs Once Caldas	Home	3.42	2.54	0.35	Marathonbet	8/19/2016 19:15	1
8/20/2016	COL	Tolima vs Junior	Draw or Away	2.38	1.85	0.29	Marathonbet	8/19/2016 19:15	1

[allen29]

What a crazy week! 2 weeks in a row with great results, that's the first time it happens.

Bankroll

Overall: 279 bets, +21.80 pts, Yield: 7.8%

Since new model: 28.49 pts from 31 bets.

[allen29]

I completely messed up the dates for the Argentine games. They are ongoing selections now

Date	League	Match	Side	Bmk.odds	My.odds	Value	Bookmaker	Timestamp	Stake
8/28/2016	ARG	Belgrano de Cordoba vs Independiente	Away	3.56	2.67	0.33	Marathonbet	8/16/2016 22:22	1
8/28/2016	ARG	Rosario Central vs Defensa y Justicia	Away	4.5	3.7	0.22	Marathonbet	8/16/2016 22:22	1
8/28/2016	ARG	Tigre vs Estudiantes La Plata	Away	3.36	2.85	0.18	Marathonbet	8/16/2016 22:22	1
8/29/2016	ARG	Lanus vs Boca Juniors	Home	2.4	2	0.20	Marathonbet	8/16/2016 22:22	1

[allen29]

Wow, how lucky must have been to receive such a personal promotion!

Anyway, selections for the future. Probably more will come for the weekend as prices will fluctuate a bit.

Date	League	Match	Side	Bmk.odds	My.odds	Value	Bookmaker	Timestamp	Stake
8/25/2016	MLS	Columbus Crew SC vs Philadelphia Union	X2	2.1	1.9	0.11	Pinnacle	8/22/2016 23:18	1
8/25/2016	MLS	Orlando City SC vs Toronto FC	2	3.06	2.61	0.17	Pinnacle	8/22/2016 23:18	1
8/27/2016	NOR	Stromsgodset vs Molde	1	2.68	2.28	0.18	Marathonbet	8/22/2016 23:18	1
8/27/2016	MXC	Tijuana vs Pumas UNAM	1	2.45	1.97	0.24	Marathonbet	8/22/2016 23:18	1
8/28/2016	MLS	Atlas vs Cruz Azul	1	3.1	2.72	0.14	Marathonbet	8/22/2016 23:18	1
8/28/2016	BRA	Fluminense FC vs Palmeiras	1	3.13	2.51	0.25	Marathonbet	8/22/2016 23:18	1
8/28/2016	MXC	Santos Laguna vs Leon	2	4.05	3.3	0.23	Marathonbet	8/22/2016 23:18	1
8/28/2016	BRA	Santos FC vs Figueirense FC	2 AH + 1.5	1.83	1.71	0.07	Pinnacle	8/22/2016 23:18	1
8/28/2016	BRA	Chapecoense AF vs Flamengo	2	2.85	2.59	0.10	Pinnacle	8/22/2016 23:18	1

 

 

Edited August 22, 2016 by allen29

[allen29]

Two new selections. Also new viz template. What you should know about this one:

• Combined filter for finding similar games into 1. Four was confusing even for me.
• Teams in the upper right chart were displayed based on their name and it was sometime confusing. Now every time home team is on the left, away team on the right.
• You can now get better insights into the players ability, particularly if they're better at scoring or assisting (left chart) and if they tend to go for a high quantity of shots or to take higher quality shots, or both (right chart). I can't wait to see Coutinho's position, I bet he's gonna be bottom right corner. I would probably name the quadrants in the future. For example players in the upper right quadrant of the first chart are something like "Creative forwards", whereas players in the lower right quadrant are more "Poacher" type.  Not all players are included, only those who have a decent number of shots or key passes to avoid one-time wonders ruining the chart.
• You will also see some capital letters in parenthesis for some players. Here is what they mean:
  • "D" - doubtful or inactive because the last time he had an event compared to his teammates is pretty unusual (could be a bench-warmer as well or a player that transferred recently),
  • "S" - suspended. Got a red card in one of the last two games. Not super accurate, as different leagues have different suspension rules and some players are suspended because other disciplinary issues.
  • "I" - (Potentially) injured. He was substituted in one of the previous 2 games because of an injury, or at least that's what it says in the commentary.

Don't take these tags very seriously though, they're not super accurate, if you want to do further research check whoscored or twitter or whatever news source.

Feedback about the charts is welcomed, what else would you like to see in there?

Date	League	Match	Side	Bmk.odds	My.odds	Value	Bookmaker	Timestamp	Stake
8/28/2016	DEN	Brondby IF vs FC Copenhagen	1	3.37	2.81	0.20	Pinnacle	8/25/2016 14:31	1
8/28/2016	MLS	Houston Dynamo vs FC Dallas	1	3.16	2.77	0.14	Marathonbet	8/25/2016 14:31	1

Edited August 26, 2016 by allen29

[wizzo1970]

looks very good, the box plots are easy to read and as it has evovled the presentation style is much better/easier to grasp.  i am going back to the first post to read through again

thank you.