I just read Stanford Wong’s book “Sharp Sports Betting” and he advises you have to take great caution using trends. He said the main issue is many just don’t have a large enough sample size for the trend to be statistically significant.
I see people posting lots of trends here that may not have enough of a sample to meet this criteria. Small samples of < 40 just aren’t enough to be relevant. Conversely if you have a large sample > 200 your trend can easily be relevant
Wong mentions 95% as the standard for data being statistically significant. That is there is only a 5% chance the next game will not meet your trend.
He likes to operate at 1/1000 or 1/10,000 error rates as he doesn’t feel 95% is reliable enough in sports betting
I will explain the math below but it might be easier is to just use his table in the book for having a 1/10,0000 chance of an error. ie your trends should be at these levels or better. So at 20 games you need a trend of ~ 95% W-L to be significant, 60 you need 75%, 100 your trend can be good at 69% , at 500 58% is ok and at 1000, 54%
I can condense his text down into some simple calcs you need to do to see if your sample is large enough.
You use 50% as the median for Wins and Losses Say your trend is CHI +3 at home is 7-3 in a sample of 10 50% would be 3-3 so you are 4 excess wins (7-3) from the 3-3 norm or median.
Take the square root of your sample size 10 or ~ 3. The is is the Standard Deviation (sd) of your sample. 2 standard deviations (95% confidence) is the most common measure of statistical significance. Wong uses ~ 3.5sd
Here that is 2 x 3 = 6 and your excess wins range is only 4. So in other words the stat isn't even 95% reliable. You would need to have 8-2 or better for that or in Wong’s case your sample wont be relevant until you have at least a size of 14.
However, if your sample is 100, and the record is 70-30, or 69-31 at 1/10,000 the data is ok.
I see people posting lots of trends here that may not have enough of a sample to meet this criteria. Small samples of < 40 just aren’t enough to be relevant. Conversely if you have a large sample > 200 your trend can easily be relevant
Wong mentions 95% as the standard for data being statistically significant. That is there is only a 5% chance the next game will not meet your trend.
He likes to operate at 1/1000 or 1/10,000 error rates as he doesn’t feel 95% is reliable enough in sports betting
I will explain the math below but it might be easier is to just use his table in the book for having a 1/10,0000 chance of an error. ie your trends should be at these levels or better. So at 20 games you need a trend of ~ 95% W-L to be significant, 60 you need 75%, 100 your trend can be good at 69% , at 500 58% is ok and at 1000, 54%
| Games | Record |
| 20 | 19-1 |
| 40 | 32-8 |
| 60 | 45-15 |
| 80 | 57-23 |
| 100 | 69-31 |
| 200 | 127-73 |
| 500 | 292-208 |
| 1000 | 539-441 |
I can condense his text down into some simple calcs you need to do to see if your sample is large enough.
You use 50% as the median for Wins and Losses Say your trend is CHI +3 at home is 7-3 in a sample of 10 50% would be 3-3 so you are 4 excess wins (7-3) from the 3-3 norm or median.
Take the square root of your sample size 10 or ~ 3. The is is the Standard Deviation (sd) of your sample. 2 standard deviations (95% confidence) is the most common measure of statistical significance. Wong uses ~ 3.5sd
Here that is 2 x 3 = 6 and your excess wins range is only 4. So in other words the stat isn't even 95% reliable. You would need to have 8-2 or better for that or in Wong’s case your sample wont be relevant until you have at least a size of 14.
However, if your sample is 100, and the record is 70-30, or 69-31 at 1/10,000 the data is ok.
