What Do You Consider Statistically Significant?

[CharlesB]

OK you statisticians, I need some help figuring out what you consider to be a statistically significant difference from a control.

To come up with a control, I ran an all burger of my database of all tracks having a minimium of 500 races. According to my figures a random pick will have an average Win% of 13% (12.82 actually) and an average ROI of .80% (.78, specifically).

Using this as a control, the question is, what would be a statistically significant difference using some angle such as rEP1 on Dirt Sprints.

It seems to me that double the random pick (around 25-26%) would be statistically significant (assuming I have enough races for that particular track, of course).

What do you think?

Are you aware of any actual rules used by statisticians for this sort of thing?

[DanG]

Charles,…

Page-4 of the March ~ 2005 newsletter…Ken explains “impact values” in detail.

This might be a good place to start.

Best of luck.

[Victor]

If I remember correctly, Ken basically said there is no such thing as statistical significance when it comes to horseracing. Big sample small sample, same difference. But what about the scratch issue? :rolleyes:

Consider this query (which does not include HTR=1 or F1=1):
Total Bets 10680
WIN PLACE SHOW
Total Amount Bet 21360.00 21360.00 21360.00
Wins 3369 5256 6441
Pct. 31.54% 49.21% 60.31%
Amount Won 21014.40 19983.70 19305.30
Profit/Loss -345.60 -1376.30 -2054.70
Pct Profit Loss -1.62% -6.44% -9.62%
Avg. Payout 6.24 3.80 3.00
Is this statistically significant?

Who knows. Horseracing is a dynamic enterprise viewed and bet on by a dynamic group. There's a lifespan to any approach you want to take.

I for one refuse to get hung up on definitions. I can count and (just maybe!) that's all the significance I need.

[Victor]

Sorry about the mistake, it does include rFR1=1 :eek:

It's so simple I'm sure most of us know the rest.

I am going to take (another!) serious look at what hurrikane talked about (scratches) when I have the time to do so. Last time I looked, it was a wash, but if anyone would post specific examples of how important it is to take this into consideration please do that.

[Victor]

This isn't for everyone, but try this: REPLACE rFR1=1 with rHTR=1 in your favorite query.

You may be pleasantly surprised!

Perfect example, the above query includes rFR1=1

Replace it with rHTR=1 and we have: Total Bets 8098
WIN PLACE SHOW
Total Amount Bet 16196.00 16196.00 16196.00
Wins 2889 4379 5239
Pct. 35.68% 54.08% 64.69%
Amount Won 15841.40 15346.00 14812.70
Profit/Loss -354.60 -850.00 -1383.30
Pct Profit Loss -2.19% -5.25% -8.54%
Avg. Payout 5.48 3.50 2.83
ok, it didn't work but don't let that stop you-- try it with YOUR favorite query!
:)
BOL