Improving the Performance of Wong Teasers

October 27, 2015 at 6:23 pm

Executive Summary: Bet Wong Teasers in NFL games with totals under 50.5 and avoid road favorites. ROI improves from 2.49% to 5.67%.

Wong Teasers Explained

A teaser bet allows the bettor to move the line in his or her favor. Teasers are most commonly offered in denominations of 6, 6.5, and 7 points, although other denominations can be found. In a 6-point teaser, the bettor moves the spread in his favor by 6 points. A teaser is a bet similar to a parlay in that the bettor must tease 2 or more teams and each leg of the teaser must win in order for the overall bet to win.

Wong teasers, also known as basic strategy teasers, were introduced in the 2001 book Sharp Sports Betting by Stanford Wong. The idea is to tease NFL spreads across the key numbers of 3 and 7. One could use a 6-point teaser to tease favorites of -7.5, -8, and -8.5 down to -1.5, -2, and -2.5, or underdogs of +1.5, +2, and +2.5 up to +7.5, +8, and +8.5, respectively. A 6.5-point teaser would be used for spreads of -9 and +1. Finally, a 7-point teaser would be used for favorites of -9.5.

In the NFL, certain margins of victory are much more common than others due to the nature of NFL scoring. Games decided by a field goal or a touchdown are much more common than games decided by 5 or 11 points. In fact the most common winning margin is 3, occurring roughly 15% of the time. The second, fourth, and fifth most common winning margins are 7, 6, and 4, occurring roughly 10%, 5%, and 5% of the time, respectively. Wong’s insight was that teasing spreads across the 3 and the 7 will catch these likely outcomes and therefore cover more often than teasing a random spread. Please note that all Wong teasers completely capture the 3 and the 7. For example, teasing a spread to -3 would not qualify; you must tease to -2.5. This is because teasers that tie are treated as a loss by many sportsbooks and we don’t want a common outcome, namely the favorite winning by 3, to result in a loss.

For a 6-point, 2-team teaser the most common odds offered are -110. To break even at -110 odds, the bettor needs to win 52.38% of bets. Therefore each individual leg of a 2 team teaser needs to win at a rate of 72.37%. I have never seen a sportsbook offer a single leg teaser. However, in order to make this analysis free from unnecessary complication, I treated each eligible game as a single bet and did not combine it with other games. A breakeven win percentage (B/E W%) of 72.37% is equivalent to odds of -261.93[i]. So, you can think of needing to beat odds of -261.93 on each leg of the teaser. For 6.5 and 7-point, 2-team teasers the odds are commonly -120 and -130, respectively. Therefore each leg of a 6.5 and 7-point teaser needs to win at a rate of 73.85% and 75.18%, or beat odds of -282.41 and -302.9.

The parameters described above are summarized in Table 1 and will be used throughout this article.

Table 1: Parameters for Wong Teasers

Spread Teaser Amt Teased Spread 2-team odds Single Leg Odds SL B/E %
-9.5 7 -2.5 -130 -302.9 0.7518
-9 6.5 -2.5 -120 -282.41 0.7385
-8.5 6 -2.5 -110 -261.93 0.7237
-8 6 -2 -110 -261.93 0.7237
-7.5 6 -1.5 -110 -261.93 0.7237
+2.5 6 +8.5 -110 -261.93 0.7237
+2 6 +8 -110 -261.93 0.7237
+1.5 6 +7.5 -110 -261.93 0.7237
+1 6.5 +7.5 -120 -282.41 0.7385


Table 2: Performance of Wong Teasers, 1990-2000

Wong analyzed all NFL games from 1990-2000. His results showed that betting these teasers was profitable in that time period. I summarized his results in Tables 2 & 3[ii]. The table shows that 6-point and 6.5-point teasers were profitable; 7-point teasers were not, which Wong attributes to sampling error. The total breakeven win percentage (B/E W%) of 72.87% is a weighted average of the breakeven percentage of each individual teaser type. To compute the ROI figures included in Table 2, I treated each teaser as a single leg bet against the single leg odds set forth in Table 1. In total, teasers returned 4.17%, or \$104.17 for every \$100 wagered.

Type W L Total W% B/E W% ROI
6-point 468 148 616 0.75974 0.7237 0.04980049
6.5-point 125 34 159 0.786164 0.7385 0.06454018
7-point 46 21 67 0.686567 0.7518 -0.0867682
Total 639 203 842 0.758907 0.72873076 0.04171677

Wong further analyzed the results of his teasers by home vs. away teams and favorites vs. underdogs. His results are summarized in Table 3. All classes were profitable except for road favorites teased 6.5 points and home favorites teased 7 points. These samples were small, 12 and 48 bets, and we can conclude little from them.

Table 3: Wong’s Results by Home/Away and Favorite/Underdog, 1990-2000

Type W L W%
6-point 468 148 0.75974
   Home Dog 134 39 0.774566
   Home Fav 130 36 0.783133
   Road Dog 172 63 0.731915
   Road Fav 32 10 0.761905
6.5-point 125 34 0.786164
   Home Dog 28 10 0.736842
   Home Fav 47 8 0.854545
   Road Dog 42 12 0.777778
   Road Fav 8 4 0.666667
7-point 46 21 0.686567
   Home Fav 30 18 0.625
   Road Fav 16 3 0.842105

I continued Wong’s work by analyzing the results of all NFL games from 2001-2014. There were 1038 games that were eligible to be Wonged. In my analysis I treated each spread as described in Table 1. I used 7-point teasers on spreads of -9.5, 6.5-point teasers on spreads of -9 and so forth. Wong teasers again proved to be profitable from 2001-2014. To compute the ROI figures, as I did with Table 2, I assumed each teaser was bet as a single leg against the single leg odds set forth in Table 1. My results are summarized in Table 3.Updated Performance of Wong Teasers

Table 4: Performance of Wong Teasers, 2001-2014

Type W L Total W% B/E W% ROI
6-point 527 191 718 0.733983 0.7237 0.0142098
6.5-point 185 54 239 0.774059 0.7385 0.04814893
7-point 64 17 81 0.790123 0.7518 0.05097636
Total 776 262 1038 0.747592 0.72930048 0.02489337

The Low Variance Hypothesis

The overall ROI decreased from 4.17% to 2.49%, mostly due to the drop in performance of six point teasers. Unlike the 1990-2000 sample, 7-point teasers were profitable from 2001-2014 which lends credence to Wong’s assertion that the unprofitability of 7-point teasers was due to a sampling error.

The theory behind these teasers is to capture the 3 and 7 because NFL games are often decided by a field goal or touchdown, or 4 or 6 points. Wong teasers take advantage of the accuracy of sportsbook lines. The more accurate the game lines, the more likely the game results will end up close to the line and the more likely the teased spread will be in the money. Therefore, I hypothesized that the performance of Wong teasers should increase in games that have low variance, i.e. games that have more predictable results.

How can we find games with low variance in relation to the betting line?

One way to reduce the variance of outcomes is to focus on games with lower expected totals, also called the Over/Under. On average, games that have lower expected point totals set by Las Vegas should have lower total scores in actuality. Lower average total scores mean that the range of outcomes is on average lower and therefore the variance of those outcomes is lower as well.

Data confirms that the lines are more accurate for games that have totals of 50 or lower. I calculated the absolute value of the difference (called mean absolute error in statistics) between the actual game results and the betting line in all NFL games, not just Wong-eligible games, from 2001-2014[iii]. The average difference between the spread and the game results for all NFL games with posted totals of 50 or below from 2001-2014 (3358 games) was 10.57 points per game. For the 218 games with totals above 50.5, the line was off by an average of 11.30 points per game. This shows that the spreads are closer to the actual results for games with totals 50 and below.

Chart 1 is a plot of the probability of winning a Wong teaser (all types) as a function of the Las Vegas Over/Under.

As you can see on the chart, there is a clear difference in the data beginning with totals over 50. The data points to the left of 50, Over/Unders below 50, are much more tightly clustered between 70% – 90% win probability. More than half of the data points for totals 50.5+ are below 70%. There were 971 Wong-eligible games with totals ≤50. The record for those games was 736 – 235 for a win rate of 75.80%. There were 67 eligible games with totals of 50.5 or greater. Those games went 40 – 27, a win rate of only 59.70%.

The 95% confidence interval for the win rate of Wong teasers in games with totals 50 or lower is from 73.11% to 78.49%[iv]. This means that there is a 95% chance that true winning percentage is between 73.11% and 78.49%. In other words, we can be almost certain that the true win rate for this subset of Wong teasers is above the minimum breakeven win rate of 72.37% for 6-point teasers.

The 95% confidence interval for the win rate of Wong teasers in games with totals of 50.5+ is between 47.95% and 71.45%, which is a fancy way of saying that these bets are very likely to be unprofitable. Based on the data, there is a 98.30%[v] chance that the true win rate of games with high totals is below the minimum breakeven mark of 72.37%. The sample size is relatively small at 67 games. So, the unprofitability of games with high totals might change as we get more data. However, the underperformance is so profound that it is likely to be significant. There is enough evidence for me to stay away from betting these games.

The difference in win rate makes sense from a theoretical standpoint. Games that have higher totals should have more scoring. As the number of scores in the game increases, the range of outcomes in relation to the spread will also increase, on average. Therefore the difference between the spread and the actual results from NFL games should also increase. Remember, for these teasers to work, we want the spreads to be accurate.

Further Analysis by Class

When comparing the performance of home teams vs. away teams and favorites vs. underdogs, the underperformance of road favorites stands out. Home underdogs, home favorites, and road underdogs were all profitable. Road favorites performed abysmally, winning at a rate of only 66.02%, well below breakeven. Table 5 shows the results broken down by home/away and favorite/underdog for all teaser amounts from 2001-2014.

Table 5: Results by Home/Away and Favorite/Underdog, 2001-2014

Type W L Win%
Home Dog 202 61 0.768061
Home Fav 254 72 0.779141
Road Dog 252 94 0.728324
Road Fav 68 35 0.660194
Grand Total 776 262 0.747592

The underperformance of road favorites held true across all teaser amounts as shown in Table 6. Larger road favorites, favorites of 9 and 9.5, performed better than favorites of 7.5, 8, and 8.5. In a small sample size of only 16 games, road favorites of 9.5 covered at a 75% rate, almost reaching the breakeven mark of 75.18%.

Table 6: Results by Teaser Amount, Home/Away, and Favorite/Underdog, 2001-2014

Row Labels Sum of W Sum of L Sum of Games Sum of W%
6-point Total 527 191 718 73.40%
Home Dog 148 53 201 73.63%
Home Fav 148 42 190 77.89%
Road Dog 189 71 260 72.69%
Road Fav 42 25 67 62.69%
 

6.5-point Total

 

185

 

54

 

239

 

77.41%

Home Dog 54 8 62 87.10%
Home Fav 54 17 71 76.06%
Road Dog 63 23 86 73.26%
Road Fav 14 6 20 70.00%
 

7-point Total

 

64

 

17

 

81

 

79.01%

Home Fav 52 13 65 80.00%
Road Fav 12 4 16 75.00%
 

Grand Total

 

776

 

262

 

1038

 

74.76%

The poor performance of road favorites with Wong-teased spreads appears to be part of a larger trend of underperformance of all large road favorites in the NFL. Road favorites of 7.5+ covered the standard, non-teased spread only 42.69% of the time (73 wins, 98 losses) in the years 2001-2014. This suggests that the marketplace for NFL betting has mispriced large road favorites. As a corollary, teasers involving road favorites of 7.5 – 9.5 are also likely mispriced and should be avoided.There were 103 Wong-eligible road favorites between 2001 and 2014. Based on this sample there is a 91.31% chance that the true win rate of this class of teasers is below the minimum breakeven win rate of 72.37%.

Improving Win Rates

The data above demonstrates that Wong teasers as a whole were profitable from 2001-2014. However, certain subsets of Wong teasers were unprofitable. Also, they were not profitable every year as shown in Chart 2.

image

The blue line in the graph shows the percentage of Wong teasers that covered in a given year. The red line shows the breakeven point for each year. The breakeven point fluctuates by year due to the relative frequency of each teaser amount. If there are more 6-point teasers in a given year, the breakeven rate would be lower, more 7-point teasers, the win rate would need to be higher. Wong teasers roughly broke even or lost in 7 out of 14 years. Ouch! Although these bets are profitable on the whole, a full season is a long time to endure a 0 or negative return.

We can improve the baseline win rate of 74.76% by not betting Wong teasers involving road favorites. Table 7 shows that the total win rate improves to 75.72% when road favorites are excluded.

Table 7: Performance Excluding Road Favorites, 2001-2014

Type W L Win%
Home Dog 202 61 76.81%
Home Fav 254 72 77.91%
Road Dog 252 94 72.83%
Grand Total 708 227 75.72%

We can further improve the performance by avoiding games with totals of 50.5 or higher. Table 8 shows that the win rate jumps to 77.05%.

Table 8: Performance Excluding Road Fav and Totals ≥50.5, 2001-2014

Type W L Win%
Home Dog 189 57 76.83%
Home Fav 242 60 80.13%
Road Dog 244 84 74.39%
Grand Total 675 201 77.05%

The year-by-year performance also improves substantially when road favorites and large totals are filtered out. Chart 3 shows that 11 out of 14 years would have been profitable.

image

Table 4 showed that the return on investment for the bet all strategy was 2.49%. When road favorites and games with high totals are filtered out, the ROI improves dramatically to 5.67%. An improvement of 3.18% may not sound like a lot, but this improved performance compounds greatly over time.

Chart 4 compares the difference in bankroll growth between betting all Wong teasers vs. betting the filtered group. The simulation begins with a \$1000 bankroll. 3.5% of the bankroll is then bet on each teaser and the running total is plotted. The change in bankroll is shown by week. Each NFL season is 17 weeks long, so each tick mark on the horizontal axis represents one season. At the end of 238 weeks, i.e. the 14 seasons between 2001 and 2014, the bet all bankroll, the red line on the chart, would have grown to \$2035.45 while betting on the filtered teasers, the blue line on the chart, would have grown the bankroll to \$4671.09.

image

I chose 3.5% as an arbitrary percentage of the bankroll to bet to keep the comparison more apples to apples. If the Kelly Criterion[vi] were applied to optimize the amount of each bet, the differences in bankroll growth would be even more dramatic. Under the Kelly Criterion, the improved win percentage of the filtered group allows a larger percentage of the bankroll to be safely bet. Therefore the return would correspondingly be much higher for the filtered group vs. the bet all group. Alternatively, the amount bet could be held constant while reducing the potential drawdown of the bankroll when using the filtered group.

Conclusion

Wong teasers have proven to be profitable for the past 25 years. Their performance has decreased since their introduction in 2001. However, their performance can be improved by filtering out road favorites and games with totals of 50.5 and higher. There is solid theoretical justification for avoiding both of these bet types, namely the general mispricing of road favorites in the NFL betting marketplace and the increasing difference between the betting line and game results as totals increase. The empirical data also bears out the unprofitability of betting Wong teasers involving road favorites and games with totals of 50.5+ with a high degree of certainty. There is a 91.31% and a 98.30% chance that the true win rate of Wong teasers involving road favorites and games with high totals is below breakeven, respectively. Filter out these bet types and your bankroll will thank you.


[i] Win probability can be converted to odds by the following formula: $\frac{(Win Probability \times 100)} {(1-Win Probability)}$

[ii] There are differences between how I present Wong’s data and how it is presented in Sharp Sports Betting. I counted ties as losses. This is the current practice in many sportsbooks and has changed since 2001.There are errors in the arithmetic in the book. I used the win/loss data provided. Wong lumps the data for 6 point teasers in with 6.5-point teasers and lumps the data for 6 and 6.5-point teasers in with the data for 7-point teasers. I teased this apart by subtraction.

[iii] I calculated the absolute error, i.e. the absolute value of the difference between the spread and actual game results: $\vert {Spread – (Favorite Final Score – Underdog Final Score)}\vert$

[iv] All random variables in this article are Bernoulli distributed. The sample sizes are large enough to assume that the Central Limit Theorem applies. The variance is the term under the radical below. The confidence intervals are calculated as follows:  ${Confidence Interval=p\pm 1.96 \sqrt {\frac 1n (p \times (1- p)}}$

Where:  $n = total # games $  $p = (# of wins) / (n) $

[v] I converted the empirical win rate to a z score based on the assumptions in iv.

[vi] An explanation of the Kelly Criterion is beyond the scope of this article. I will say that it is a rule for optimal bet sizing.