I have geared a spreadsheet to generate predictions for a Sports League based on Least Squares Regression of previous results and Poisson Random Variables (Team Scores that seem to fit as Random Variables).

I am trying to refine the data, to remove or identify results that aren't as likely as predicted. One factor I am looking at considering is streaks (e.g. Team A has won its last two games, Team B has lost its last two, hence I can decide that Team A is more likely to win a close game).

Columns I have in the table are:

A - Round (think of it as week 1, week 2 etc.)
B - Date (I'm not sure if this is formatted as a date)
C - Time of Game
D - Home Team
E - Home Goals
F - Away Team
G - Away Goals
H - Home Margin of Victory (E - G so it will be negative for a loss)
The other columns on the table hold other irrelevant data

I'm thinking I could probably do this with a COUNTIF formula, and may need to generate a winning team cell e.g. (=IF(H31>0,D31,F31) first, but I'm really not quite sure of where to go with this (or would I need to generate some kind of lookup table first - to almost count the current streak).

I would really appreciate any advice anyone may be able to offer.

Thanks,