Sportpunter's MLB 2006 Report Card

Welcome to the Sportpunter Major League Baseball Report Card. Here we will look closely at how we went in 2006 and compare this to previous years and well as forecast how we think we will go in 2006 and beyond. We will also look at in depth data in order to maximise your profits using the proper money management. Please feel free to comment on the analysis in the forum thread shown here:

The report will be split up into the following contents:

1. Betting History of MLB in 2006
2. Analysis into different probs/odds/overlays
3. Optimal Kelly Fractional
4. 2006 Bank Analysis
5. 2006 and Beyond

1. Betting History of MLB in 2006

The MLB Baseball Season in 2006 was once more a huge success. We gained significantly on head to head matches as well as totals betting. Figures below show that a very significant profit was made betting head to heads that was comparable with previous years. Thus indicating, that the model is standing up to the test of time.

For totals betting, a great profit was achieved, however significantly less than previous years. However subscribers will remember the large amount of overs at the start of the season, which compared to previous years was an unexplained irregularity. From may onwards when it seems that the issue had been resolved we managed to make $7633 and return 8.7% ROI. This is slightly less than in 2005, however greater than the two years before this and is still a very admirable result.

Previous Years Betting History:

Totals Betting History:

Graphs of the two types of betting for the year are shown below. ***

The graphs show that for head to head betting, we achieved large profits and losses in a cyclic type pattern, however the overall pattern was increasing. From mid year until the end of the season, we recorded constant profits. 2005 saw us record good profits in the last couple of months of the year as well.

Totals betting saw a decrease at the start of the year, but saw constant profits for the remainder of the season despite a small hiccup mid season.

2. Analysis into different probs/odds/overlays

Shown below is the results for MLB given in three tables for different probabilities, odds and overlays. All analysis on totals betting is from May onwards.

H2H betting:

As shown from the link above, we failed to make a profit from probabilities less than 45%. Our model performed amazingly well for overlays of between 5% and 17.5% gaining a greater %ROI than at the very best expected (this shows how much gambling can vary - eg in the 5% to 7.5% bracket we would expect at most 6.25% ROI, however re gained 3 times this amount). Larger overlays produced greater profits as expected.

Totals Betting :

Interestingly high and low probabilities failed to make a profit, however the sample is far to small to make any long term judgments. Profit generally increased with greater overlays. The is no significantly beneficial or detrimental line to bet at.

Then looking at just the Overs, a greater profit was made and increased %ROI. A small amount of larger overlay bets means it is hard to determine whether there is a significant relationship between overlay size and %ROI. Once again no pattern emerges in specific run lines. The case for betting Unders held similar patterns although not as strong a profit. It is noted that betting under lines of 8.5 provided only 1% ROI, however we should not dwell too much on this given the small sample size.

What is clear is that the MLB model, both h2h and totals performs very well in all situations and categories.

3. Optimal Kelly Fractional

When looking at the optimal Kelly Fractional we have to work out the expected profits given the overlays. Please note that any calculated optimal Kelly fractional is dependant on the previous years data and is by no reason a Kelly fractional that will increase profits the most in future years. It is purely a means of what would have been the best Kelly Fractional in the past, and quite possibly might be in the future. In 2005 we calculated that for h2h betting, the optimal kelly to use is 3.2 and 1.5 for totals betting.

Expected profits in 2006 for head to head betting on favourites was $27,216. This is a %ROI of 21.9%, which is almost exactly the same as last year. As the actual %ROI was 8.4% ROI, this equals therefore an optimal Kelly fractional of 21.9/8.4 = 2.6. Or in other words 1/2.6th Kelly fractional.

Expected profits in 2006 for totals betting since May was $12,778 which gives a %ROI of 12.7%. Given the actual %ROI was 8.66% this suggests an optimal Kelly fractional of 12.7/8.6 = 1.5. Or in other words 1/1.5th Kelly fractional, which is the same as last year again. This is incredibly low and we have to discuss the meaning behind this.

This shows that for two years in a row we have provided some very good totals betting results. The consistency of totals betting results even surprises me. I, of course do not recommend betting anywhere near the 2.6 or 1.5 kelly recommended but would suggest somewhere around 1/4th to 1/16th kelly. Around 1/8th would probably be ideal. This analysis does however suggest that the totals betting is more accurate and consistent than the head to head betting. Hence some people might want to have a slightly lower kelly factorial for totals than head to head, but any difference should be marginal.

Keeping in mind that this analysis for totals does not include the first month of last years season where just about every suggested bet was for an under, and scores were all going over. Sportpunter have adjusted the way that we provide the predictions so that this early season variability will not hinder our bets.

4. 2006 Bank Analysis

For head to head betting, the following is a graph of the bank for different constant Kelly's.

As one can see the lower the kelly factorial, the greater profits. But also with this comes the greater variability and risk. For example between the dates 18/6/6 and the 2/7/6 the model lost a significant amount. 1/3rd constant kelly would have lost 50% of their bank, 1/4th kelly 43% of their bank, 1/6th 35%, 1/8th 30% and 1/12th 23%. Hence higher kelly's, whilst not gaining as much money had less variability and risk.

The same applies when looking at moving banks as shown below.

Lower kelly's resulted in great profits but more variability. During the dates listed above, 1/3rd moving kelly lost 75% of their bank, 1/4th kelly lost 64%, 1/6th lost 50%, 1/8th lost 40% and 1/12th lost 28%. Interestingly the greater the kelly factorial, the more comparable the end bank results are for moving and constant kelly. However Moving Kelly always was ahead of it's constant counterpart.

Similar patterns are shown below for totals betting. Whilst not as consistent as last year, similar profits were made.

The constant profits as shown above in the constant Kelly bank obviously show the consistency in the model in 2006. The difference between the constant Kelly graph and the moving bank is very dramatic. The surreal bank of the moving bank is indeed unbelievably large, but this outlines the advantage of using a moving bank over a constant bank should you be betting with profits. The moving Kelly graph is based on an adjusted bank at the start of every week.

Alternatively we can also look at keeping one sole bank for both h2h and totals betting. Doing this (and only counting totals since May) the results for different bank kelly's are shown below for constant and moving banks.

Similarly, it shows that whilst higher kelly's showed more profit the potential for bankrupcy and greater variability is large, with 1/3rd constant kelly losing 33% of the bank at one stage down to 1/12th constant kelly losing 19% at most last year. What is also clear is that moving banks made more money in every instance than constant kelly, however the variability is massive. 1/3rd moving kelly during the bleak mid season lost a staggering 79% of the bank going from $6500 down to just over $2000. It managed to regain this amount and some, but realistically who wants to put themselves at risk of losting 4/5th of your bank.

For want of comparison, 1/4 moving kelly lost 68% in the same period of time, 1/6 moving lost 52%, 1/8th lost 41% and 1/12th lost 30%.

The layoff between potential risk and potential profits is of course up to you, but if unsure I would recommend the higher kelly factorials between 1/8th and 1/12th.

5. 2007 and Beyond

Another great year for Baseball betting in 2006. And there is no reason why the success will not continue in 2007. Previous years suggest that 2006 was on relatively equal par with that of previous years. Given changes to the totals model, we should not see again the terrible start that we had in 2006.

Later releasing of data should not play a part on the model's success and will only allow punters to bet more and more variety of places to bet at. Our masses should also not move the prices as much as last year. We are also going to have strict upload times to make the information easier to utilise.

Sportpunter’s MLB model is clearly a class above the rest, having unbiased mathematically based predictions independent of media hype. The model maintains memory of all games, and establishes the importance of variables and their influence over the game; for example weather, and wind conditions.

We hope that if you enjoyed the fruits of Sportpunter’s baseball in 2006 that you would not hesitate to sign up again in 2007. And if you are new to Sportpunter’s baseball model, then we recommend that you stop looking at the profits we have made in the past and present (as shown above), and join in to also reap the rewards.

Best of Punting Luck
Jonathan Lowe
www.sportpunter.com