Monte Carlo Tennis Prediction Challenge ATP

[britbox]

A fun game where you score points for predicting the match results in the tennis tournaments. You pick the winners at the beginning of the tournament.

This takes as little as 10 minutes per bracket to play

Instructions

Video Tutorial Here:



Tennis Draw Challenge Instructions

or for a larger view:

https://discuss.tennis/tennisforum/index.php?media/tennis-draw-challenge-instructions-youtube.25/

In a nutshell...

If you aren't a member of the discuss.tennis website (the one you're reading - sign up - top right and then we can track your score), then...

1. Goto http://www.tennisdrawchallenge.com/ and register a new account (if you haven't got an account already). Try and use the same username as you use on this (the discuss.tennis) website.

2. When you are registered, click the GROUP menu option (at the top) and join the Discuss.Tennis Group (we're near the top).

3. Then click BRACKET menu option at the top and you'll see the tournaments for either WTA or ATP. Click In the second column, you will see a "Make your picks now" link. Click on this to make your picks. You can toggle between ATP and WTA tournament listing by clicking the "WTA" and "ATP" buttons at the top.

You're playing for a place in the Hall of Fame! - https://discuss.tennis/tennisforum/pages/hall-of-fame/

 

[britbox]

The Bracket will be open on the Draw Challenge website shortly after the draw.

 

[mrzz]

I've been away on the holidays -- no internet there (envy me, robots) -- and did not make back before the start of the tournament. Won´t miss the next one.

 

[mcekovic]

Monte Carlo tournament simulation based on neural network match prediction: http://tennis-strangeforest.rhcloud.com/inProgressEventForecast?inProgressEventId=9

 

[Moxie]

Monte Carlo tournament simulation based on neural network match prediction: http://tennis-strangeforest.rhcloud.com/inProgressEventForecast?inProgressEventId=9

Very curious how you (or whomever) came up with such specific percentage chances by rounds and quarters. And welcome, mcekovic!

 

[Moxie]

I was away for the weekend, too, and thought I had until Sunday night. Bummer. Good luck, all.

 

[britbox]

Same here. Was away camping - couldn't get online and missed it by a few hours.

 

[Moxie]

Same here. Was away camping - couldn't get online and missed it by a few hours.

You, Mrzz and I will just have to sit by and smugly insist that we would have made all the correct choices, if we'd only had the chance.

 

[britbox]

You, Mrzz and I will just have to sit by and smugly insist that we would have made all the correct choices, if we'd only had the chance.

I was quietly confident of getting 100% on the predictions this time around. I had you and Mrzz at around 7% and 2% respectively. I think this is probably the safest most realistic option to follow when we tally up the final results.

 

[Moxie]

I was quietly confident of getting 100% on the predictions this time around. I had you and Mrzz at around 7% and 2% respectively. I think this is probably the safest most realistic option to follow when we tally up the final results.

I'll see you at the next draw challenge, Mr. 100%.

 

[mcekovic]

Very curious how you (or whomever) came up with such specific percentage chances by rounds and quarters. And welcome, mcekovic!

At the core of the tournament simulation is the single Match prediction which is based on the Neural Network Algorithm with ~50 neurons for different features about players like Elo Rating, Surface Elo Rating, ATP Ranking, ATP Points, Head-to-Head ratios and Winning Percentages varied by surface, tournament level, round, recency, match or set ratios, vs rank, vs hand, vs backhand...
Neural Network is then trained on historical data (last ten years) for highest prediction rates to determine optimal feature weights.
Elo ratings, total and by surface are main contributors to the prediction, following by head to head (total and by surface), recent form and vs hand (vs lefty or vs righty winning percentage). Then come other factors, like tournament level (GS, Masters, ATP 500, 250, etc...), round, vs rank (winning percentage vs players of similar ranking), vs backhand (winning pct vs players with one- or two-handed backhand)...

Prediction for the next round is then obvious, it is the probability of the very next match outcome.
Prediction in future rounds is then based on who are the potential rivals in that round. Usually, for the first round after the next (two rounds in advance), there are 2 potential rivals, for three rounds in advance 4 potential rivals, for N rounds in advance 2 ^ (N-1) potential rivals.
All of them have probabilities to reach the previous round. Then match prediction is done for each of the potential rivals, and total probability is sum of products of these two probability sets.
As the tournament goes on, some matches are completed, then winner probability is obviously set to 100% and loser to 0%.
Prediction is updated every hour, so in case of big upsets, you can see probability of other favorites increasing, while in case tournament goes with the seeds winning, you can see probability of favorites stagnating, even they progress in the next rounds (as they get guaranteed harder rivals, i.e. seeds), or even falling down.

 

[mrzz]

I was quietly confident of getting 100% on the predictions this time around. I had you and Mrzz at around 7% and 2% respectively. I think this is probably the safest most realistic option to follow when we tally up the final results.

Classical mistake of promising results. On the other hand now all I need is to do better than 2%. But I won´t promise nothing.

 

[britbox]

Congrat @Tyler D for winning Monte Carlo Draw Challenge.

http://www.tennisdrawchallenge.com/group/standings/41/atp/bracket