Newhavensoft forum

**Kervon** · Newbie

Hi,

I do remember spotting this question being asked somewhere in the forum, but I wish to have a discussion regarding the optimal depreciation period for the historical data.

The question lies is : Is 900 days the optimal DP?

I've actually read up quite a number of sources which suggested that historical data for any predictions in soccer should be kept at a maximum of 'last 2 seasons' Any more than that would cause an a decrease in stability.

Has anyone tried alternative DP? I've personally tried 600 days, the difference in the ratings didn't seem make any significant difference, but the correct score odds tend to change quite bit.

Any opinions?

-----
"Winning can only come after you have instilled the disciplines"

**Flow** · Newbie

I think 900 days is the best period for Odds Wizard but only for soccer (and Ice Hockey).
Why ? Beacuse Alex said that

Only for soccer ? Why ? Because I've tried other sports like Basket-Ball (French, Italian, Spanish), Rugby (French, English), Handball (French) and the home team is always the favorite.
I've tried to compute the last team against the first and the last has the advantage even if they have won only 3 and lost 20 and the first team has won 18 and lost 5 games.

Is Odds Wizard also right for other sports (good for soccer and ice hockey) like basketball, handball, rugby, ... ?

(Edited by: 19 March, 2006 - 04:35:14)

**Kervon** · Newbie

Hi Flow,

Well, I believe I need more solid and confident answers than just "Because Alex said it"... No offence meant, but I'm in need of a serious reply, not one to gain Alex's attention.

Anyway, although I don't quite understand what you're trying to say. But I guess you're trying to explain that in other sports the home team always has the advantage. Well, I thought that has always been the case. No matter which sports, home advantage has always been there, it's proven and the advantage cannot be disregarded.

The poorest performing team in a league will still have a home advantage over an away team that is first in the league. Do you get the picture? OMG, I really don't know how to explain it further...

-----
"Winning can only come after you have instilled the disciplines"

**admin** · Administrator

This topic is very close to that about 'How do we know the OW odds are fair odds?' http://www.newhavensoft.net/forum/topic.php?forum=1&topic=19.

By varying the depreciation period (DP) it is possible to achieve somewhat better values of Total Success rate (in Performance analysis tool). At the same time, one should remember that computed 1X2 distribution (Relative rate in column 3) must be as close to actual (Relative rate in column 5) as possible. Otherwise computed odds cannot be regarded as fair odds.

800-900-1000 days are the best values for the vast majority of soccer leagues. However the 'memory' of international competitions, especially World Cup, is much stronger, and therefore requires longer DP.

**ColdFish** · Newbie

I have tried experimenting with changing the depreciation period for the Spanish leagues. Increasing the depreciation period for the first division has little or no effect in performance of forecast accuracy, decreasing the period below 900 results in a decline in forecast accuracy.

With regard to the second division changing the depreciation period from 900 up or down results in no significant change in accuracy. The figure stays at around 101.5.

I suspect this effect is due to the significant change in teams within the lower division, and the makeup of players. Divisions with historical long standing teams are more forecastable.

Why 900 seems to work well I do not know.

-----
ColdFish