Jump to content
Download the PL Forum App!
** ELO Ratings are now back **

Tennis Forecasting system

dalkent

By dalkent
in Betting Systems & Strategy

 Share

Recommended Posts

Matthew

Matthew

Posted
Posted

Re: Tennis Forecasting system Hi dalkent, I haven't read anywhere in the thread so far where you talk about the modelling process, the software you've used and the type of model you've implemented. I've got a number of macros, data and ideas for tennis betting at my disposal. Have you a dropbox account? Would happily share some of it with you.

Link to comment
Share on other sites
  • Replies 323
  • Created
  • Last Reply

Top Posters In This Topic

Popular Days

Top Posters In This Topic

Popular Days

dalkent

dalkent

Posted
  • Author
Posted

Re: Tennis Forecasting system Hi Matthew, Wondered when anybody would ask about the models. There are currently 5 models in total. 3 for men's and 2 for women's tennis. All the models are slight variations on each other. The initial datasets were formed in excel where i had to write a number of VBA functions to calculate the model variables. I started with many and narrowed them down during the modelling process. I had about 50 different variables which were mainly variations on number of wins on each court type over different time periods. Also players rank was used as a variable. Once the datasets were formed I build the model using the stats package 'R'. The models are logistic regressions models which are convenient since they output a probability of win/lose and therefore can easily be converted to odds. The reason for multiple models is that for many players sufficient historical data is not available. The most advanced model used Rank and Variables of variations of number of wins per court type over the last 6 months. If enough data is not available then the second model is invoked which uses data from 13 years. i.e. player win rate over the last 13 years. If there is even not enough data for this then a simple base model just using players rank is produced. Every time the model runs a dataset covering the last 13 years is queried to populate the variables and produce the results. My next steps are to add head to head variables although this will probably only be useful when the top players play each other. I am also looking to add in a ELO type ranking system like that used in Chess. I would prefer o remove the players Rank from the model if possible. I would be interested to know your thoughts on tennis betting. I do have a dropbox account but must admit no idea how to share folders etc. I will look into it. Cheers

Link to comment
Share on other sites
dalkent

dalkent

Posted
  • Author
Posted

Re: Tennis Forecasting system Results for trading have been relatively good although could have been better if greed had not got the better of me. A simple process i use is as follows. 1- Bet 3% of bank on odds that have greater then 10% value when compared to model fair odds. 2 -Setup reverse trade to take 20% profit (sometimes an early break can give much greater return 3 - Try and put a stop loss in place of 20-30% loss (must admit haven't stuck religiously to this to my cost) 4- Try and get out of trade before end of first set as large swing in odds at end of set. Current positions is Bets 13 S/R 69.2% ROI 9.2% Current Return on Bank 3.26% Obviously i can't have my results validated on here since I can't post before trades etc. So i appologise about that but thought interesting to post my approach.

Link to comment
Share on other sites
Matthew

Matthew

Posted
Posted

Re: Tennis Forecasting system

Hi Matthew, Wondered when anybody would ask about the models. There are currently 5 models in total. 3 for men's and 2 for women's tennis. All the models are slight variations on each other. The initial datasets were formed in excel where i had to write a number of VBA functions to calculate the model variables. I started with many and narrowed them down during the modelling process. I had about 50 different variables which were mainly variations on number of wins on each court type over different time periods. Also players rank was used as a variable. Once the datasets were formed I build the model using the stats package 'R'. The models are logistic regressions models which are convenient since they output a probability of win/lose and therefore can easily be converted to odds. The reason for multiple models is that for many players sufficient historical data is not available. The most advanced model used Rank and Variables of variations of number of wins per court type over the last 6 months. If enough data is not available then the second model is invoked which uses data from 13 years. i.e. player win rate over the last 13 years. If there is even not enough data for this then a simple base model just using players rank is produced. Every time the model runs a dataset covering the last 13 years is queried to populate the variables and produce the results. My next steps are to add head to head variables although this will probably only be useful when the top players play each other. I am also looking to add in a ELO type ranking system like that used in Chess. I would prefer o remove the players Rank from the model if possible. I would be interested to know your thoughts on tennis betting. I do have a dropbox account but must admit no idea how to share folders etc. I will look into it. Cheers
Thanks for the breakdown. Not often you see people applying statistical techniques here on PL, so when they do I like to have a chat about it as Sports, and the proliferation of data, lends itself quite nicely to a approach such as yours. Do you have any output from your Logistic Regression model builds? Model performance on the build, test, validate samples? Gains curves, parameter estimates....that sort of thing? A couple of extra things I'd suggest you take a look at, and I think might already be on your radar are: Performance in the event the previous year - what ranking points is the player protecting - a major motivation Time on court - a few metrics here, you've got intra tournament, in the past 30 days, in the past year - players tend to be a little more protective of themselves as the season wears on. Time off court - has the player taken an extended break - could they therefore be returning from injury - for example, how would your model have dealt with Nadal's long absence? Grit - is the player a fighter, some players lose interest. Performance from a set down in 3 set matches and from 1 or 2 sets down in 5 sets matches Performace v's left handed players Performace v's certain types of players (you could cluster players based on some key metrics so you can assess how people play against baseline players or serve and volley or powerful servers) I have a dataset of prev tournament performance and match stats (which includes time on court) that I would happily share. On dropbox, if you go to sharing and create a new shared folder I'll put them dataset in there for you. If you have used tennis-data.co.uk datasets then it should match in nicely. PM me when you've done this and I'll tell you my dropbox details. Good luck Matt
Link to comment
Share on other sites
welshpaddy

welshpaddy

Posted
Posted

Re: Tennis Forecasting system Dalkent - After I posted Federer I found an error in my new excel work sheet, it means I needed to get 1.56 not 1.53 but to late to now so i'll stick with it. Also, since the 22nd I started noting which model each selection came from and Advanced 2 Model is doing well, but early stages so far. I've also noticed the womens model is not as good but you did warn me of that. Maybe because there is a lot more break of serve in the womens game. I'll keep on till the end of the season and see how it goes. Also, I really fancy Janowicz to beat Ferrer today, i've taken 3.5 but may trade out after the first set as Ferrer seems to win from behind alot. Good luck

Link to comment
Share on other sites
dalkent

dalkent

Posted
  • Author
Posted

Re: Tennis Forecasting system New day. New games, New stress Wawrinka to be Almagro £3 at 1.6 - Win Ferrer to beat Simon £3 at 1.4 - Win Djokovic to beat Isner £3 at 1.2 - Win Del Potro to beat Dimitrov £3 at 1.4 - Win On a good run. 7 wins in a row

Link to comment
Share on other sites

Join the conversation

You can post now and register later. If you have an account, sign in now to post with your account.
Note: Your post will require moderator approval before it will be visible.

Guest
Reply to this topic...

×   Pasted as rich text.   Paste as plain text instead

  Only 75 emoji are allowed.

×   Your link has been automatically embedded.   Display as a link instead

×   Your previous content has been restored.   Clear editor

×   You cannot paste images directly. Upload or insert images from URL.

×
 Share
Go to topic listing
×
×
  • Create New...