2 Suggestions: Complete historical data and "Command" rating

[hockeyman001]

Complete Historical Data
I realize what an enormous undertaking this would be, but considering the game does seem to factor in splits and batter vs pitcher when simulating an at bat, it would seem to significantly add to realism.

This would incorporate historical statistics into splits such as players who hit significantly better at home than on the road (which may not be accurately modeled when loading a historical ballpark when starting a previous season), while still leaving enough variance that the results don't seem scripted. Splits seem to have a significant impact on player performance in season (I am using 70 ratings / 15 this year / 10 last year / 5 two years with development engine on - no historic stats - for a heavily scouting based game). For example certain players have hit many of their homers at home by chance and even with that AI split, they tend to play out that way...they are studs at home and singles hitters on the road. This particular feature would probably help normalize that a bit, as well as provide more data for consumption. Also as part of this: adding first half/second half splits, and 2 out/RISP stats.

I realize this would be a time and game memory suck, but perhaps it could be added in working backwards from present day stats until the splits are no longer available, or perhaps just for key splits. I believe it would add greater realism (AKA Trot Nixon owns Roger Clemens, and A-Rod isn't so great with 2 outs and RISP).

I realize this is modeled into the ratings system (especially "vs left and right" split ratings), but in my experience players who get off to a hot small sample size start can often tend to exceed expectations for very prolonged periods of time, regardless of ratings, and I'm saying this using a very ratings heavy AI distribution. There seems to be a point where the stats factor into the calculation enough to become self fulfilling, without the game seeing fit to adjust the underlying ratings.

Command Rating
This has been brought up before. An additional rating to control/movement/stuff specifically for home run avoidance. Obviously it would be counteracted somewhat by HR/FB normalization, same as movement is counteracted somewhat by BABIP normalization. However it's often difficult to accurately model pitchers like Michael Pineda for example in game. High strikeout, high home run, yet low .BAA (in good years) type pitchers. It would just be another tool IMO to generate more accurate player projections in game, and to somewhat control HR/FB without having to use extreme values in movement. On the other end of the spectrum, guys like Tom Glavine or Jamie Moyer tend to be pinatas despite the fact that their "movement" (using the in game term) was enough to let them consistently be very good despite weaker "stuff".

Again, I'm not talking about small sample size issues, just accurate player modeling issues...this has nothing to do with in game results, just the projected normalized stats the game generates which are a basis for simulated results.

Just my thoughts. Thanks for reading the whole thing!

[Rain King]

Delete