A formula too good to be true?

[rjp]

I use recent results (adjusted to remove home court bias) to determine what the odds for a given matchup should be if the teams were to play at the same level as they've been over their recent results.

There are plenty of reasons why a team is likely to play better (or worse) than what recent history suggests. Injuries, motivation, etc. You'll need a much larger sample size to know if you're on to something.

[Dan82]

Too much trouble? Not tough tough if you have the data from all the games in a spreadsheet. I have been using 2 numbers...overall and last 5 games. Average the two together or weight them accordingly. Maybe even disregard the first few weeks of the season at the end of the year. How is that difficult?

[NoChance]

If it is working then give it a go. It may be easier for you then it was for me. For me is was just too much to keep up with. Just keep in mind some of the other factors. Such as, what numbers will you use when Shaq returns? What about other team changes and injuries? Do you figure home and away? How about back to back nights? Tons of factors that go beyond numbers that makes this difficult.

[Dan82]

I hear you chance. Tonight, it will be a pain to input all the data...too many games. But it helps when your an unemployed poker player lol. Did you have any success quantifying home court advantage and back to back games? I have been experimenting using a number between 1 & 3 for both of them.

[iggymcfly]

I don't know about for back-to-backs, but 3 is considered pretty standard for home advantage in basketball.

[NoChance]

That is the part that was too difficult for me and thus not worth my time. THinking more about it, I kink of went through stages.

Stage 1 - Hey I found a formula that works! (early in the season)

Stage 2 - Not working so good anymore. I'll make an adjustment and throw out early games as teams have changed. (early-mid season)

Stage 3 - This is bad! I start going opposite because I figure my data contains too much irrelevant data and the true line must be correct if my data is this far off and I keep losing. (mid-late season)

Stage 4 - I give up. No way to quantify injuries. No way to quantify home/away stats for example if you take the last 10 home games those last 3 or 4 games could be from 8 weeks ago. Thus, not enough data for the numbers to be accurate. No way to quantify injuries. Take Shaq for example. The numbers just don't skew Miami's stats but also EVERY team they played while he is out.

--------

Anyway, that is how it went for me. It may or may not go that way for you. All the more power to you if you can find a way to make it work. If anything we are rooting for you to find the golden formula! We would all be very happy if you did and piggy-back off of you. For me, not worth it.

[HeadWaiterSports]

I think your "formula" is a good starting point but you have to factor in a lot of other elements. If you have a team that has just had a long home stand, those numbers are going to be in their favor for the beginning of a road trip and may not be a true indicator of where they stand. Also, you are not taking into account a back to back game. There is a lot of credence in how well a team played the night before, etc... I would recommend that you use your new system to get a base of games and then start digging into them to come up with your actual plays. A good handicapper never looks at the games from just one aspect. A game should be a positive from just about every angle you have and then you make your play.

[Dan82]

No picks for tonight

12/01/06

Charlotte (+6.5) @Washington
Clevland @ Atlanta (+2.5)
Chicago @ New Orleans (+4)
New York (+8) @ Detroit

[smbruin22]

a guy in my office did the most simple power ratings and then adjusted for home/away (but probably same for each game) and did really, really well in office pools for years..

sometimes i think things are made too complex.... although i guess in fairness he applied a little interpretation as most power ratings will give too large a spread.

[awieise]

On page 132 of Stanford Wong's book "Sharp Sports Betting", there is a table that you can use to determine whether or not your candidate is truly a winning system, as opposed to just a lucky fluke.

For a sample size of 80, for instance:
if your record is 51-29, you can achieve a record this good 1 out of 100 times just by guessing, so this isn't so good.
if your record is 54-26, you can achieve a record this good 1 time out of 1000 just by guessing, so this is "probably a system that will make you money".
if your record is 57-23, you can achieve a record this good only 1 time out of 10,000 just by guessing or luck, so "you truly have a winning system".

And of course the numbers and percentages change as the sample size changes.