How the Simulation Works

[Rockies GM (Dan)]

There have been questions as to how the program figures things out, whether any bias on my end goes into it, etc. So, to keep things as transparent as possible, here is how it works.

Let's take one of the more significant matches this week - Minnesota vs Oakland. Here's how the two teams stack up in Z-Scores based on last week's stats

Through equations based on historical correlation between Z-Score and category winning%, the computer spits out odds for each team in each stat, and we get this:

Then, I compare each of the stats through another algorithm, designed to figure out statistical odds of an event happening, given odds of two different outcomes (in this case, given the odds that one team will win vs the odds that the other team will win). With that:

Stat	Favorite	Odds
Runs	Minnesota	62.61%
Home Runs	Minnesota	51.96%
RBI	Oakland	50.58%
SB	Oakland	58%
Average	Minnesota	68.2%
OPS	Minnesota	69.05%
Wins	Minnesota	56.35%
Saves	Oakland	71.76%
Holds	Minnesota	51.88%
Strikeouts	Minnesota	56.7%
ERA	Minnesota	53.31%
WHIP	Minnesota	51.73%

I then generate a series of 12 random numbers to compare against these percentages (all compared against Minnesota's in this case):

R - 6261>5254 - Minnesota wins
HR - 5196<6724 - Oakland wins
RBI - 4942<5093 - Oakland wins
SB - 4200>3869 - Minnesota wins
AVG - 6820>4829 - Minnesota wins
OPS - 6905<9161 - Oakland wins
W - 5635<9630 - Oakland wins
SV - 2824>1497 - Minnesota wins
K - 5670>3996 - Minnesota wins
HLD - 5188<9350 - Oakland wins
ERA - 5331<5823 - Oakland wins
WHIP - 5173<6193 - Oakland wins

Oakland wins 7-5

I have a tie-breaker system in case of a tie, and at the moment, the system is unable to produce a true tie, which doesn't bother me, due to the paucity of them in real life.

Something I may opt to add in the future is a way in which to allow the statistical category to tie. But the research on that would have to be extensive - the odds of one Z-Score tying another Z-Score based on history.

And that simulation above gets simulated 100,000 times in order to account for statistical anomalies.

If you're wondering, Minnesota has a 60.81% chance of winning the matchup, according to those numbers

[Rockies GM (Dan)]

Something else I may look to add eventually - a way to link AVG and OPS, along with ERA and WHIP. But that's also on the backburner - it's fairly normal for one team to win one of those stats and the other team the other

[Former Twins GM (Robin)]

Wow. Great info Dan. 👍