A projection system for OOTP 26

[Guthrien]

Originally Posted by jaa36

well, it's a fairly well-documented real-life effect

That's amazing, thank you for sharing that. I was so convinced of my bias on that I simply took it as a given you were discussing a game factor. I have to admit, I never ran into that one but it's interesting. The injury path is clear but this is more symptomatic. Ah, baseball.

[trence]

Has anyone had an issue after many seasons of the tool working? I tried both V1 and V2 and they fail. I'm getting this error on V1. First season I've had issues with it suddenly crashing out. Wondering if somehow it's exporting bad CSVs all of a sudden
--------------------------------------------------------------------------------------------------

ValueError Traceback (most recent call last)
<ipython-input-220-1929e306d16c> in <cell line: 0>()
27
28 # Apply function to dataframe
---> 29 merged_df[['has_pos', 'field']] = merged_df.apply(determine_positions, axis=1)
30

2 frames
/usr/local/lib/python3.11/dist-packages/pandas/core/indexers/utils.py in check_key_length(columns, key, value)
388 if columns.is_unique:
389 if len(value.columns) != len(key):
--> 390 raise ValueError("Columns must be same length as key")
391 else:
392 # Missing keys in columns are represented as -1

ValueError: Columns must be same length as key


This is from the section

# calculated modified best position based on fielding ratings - i.e. whether a hitter 'has a position' or is just a 1b/dh

[squirrel]

Originally Posted by trence

Has anyone had an issue after many seasons of the tool working? I tried both V1 and V2 and they fail. I'm getting this error on V1. First season I've had issues with it suddenly crashing out. Wondering if somehow it's exporting bad CSVs all of a sudden
--------------------------------------------------------------------------------------------------

ValueError Traceback (most recent call last)
<ipython-input-220-1929e306d16c> in <cell line: 0>()
27
28 # Apply function to dataframe
---> 29 merged_df[['has_pos', 'field']] = merged_df.apply(determine_positions, axis=1)
30

2 frames
/usr/local/lib/python3.11/dist-packages/pandas/core/indexers/utils.py in check_key_length(columns, key, value)
388 if columns.is_unique:
389 if len(value.columns) != len(key):
--> 390 raise ValueError("Columns must be same length as key")
391 else:
392 # Missing keys in columns are represented as -1

ValueError: Columns must be same length as key


This is from the section

# calculated modified best position based on fielding ratings - i.e. whether a hitter 'has a position' or is just a 1b/dh

Hmm...tough one to help with there just going by that error message. Most simple place to start is to check that you've updated the ID of your scouting director to be the latest one.

This is on v1, right i.e. not the latest stuff I put out?

[trence]

Originally Posted by squirrel

Hmm...tough one to help with there just going by that error message. Most simple place to start is to check that you've updated the ID of your scouting director to be the latest one.

This is on v1, right i.e. not the latest stuff I put out?

Oddly enough, I was using both v1 and v2 just to mess around. Then during a working save, they both started failing. One is run locally, one is on Google Colab.

Manager ID and all that was setup and working fine for years on the franchise, then just stopped one year on V1 and V2.

[Guthrien]

Everyone still playing around with this? Any discoveries on your part @squirrel?

I still very much enjoy it, especially around the trade deadline. It's absolutely a contrast machine and not a stat projection, I'm not certain what it would take to drag that part closer. I used to think OOTP calculator (with the simpler stat basis) wasn't horrible at times, but it served the same purpose - comparison.

There are values in pitching that it seems too biased towards. My example will make more sense to any Tiger fans among you, but pitchers like Jack Flaherty, high stuff, low HRA, are punished to the point you'll always be told he's a horrible cancer to your staff instead of the 2nd or 3rd best starter he will always be. Meanwhile, mediocre talents with a groundball tendency will often be projected as your no argument #2 starter (we have Skubal, nothing gets him wrong). Batting projections are far less chaotic.

[squirrel]

No new discoveries here...just seems to work pretty well now as I wanted it to. I don't think I'm going to need to tweak it any more

[sabres12]

I am not well versed in Github so figured here would be the next best place for this. I made some changes to my local copy of this and wanted to share. I have been using this exclusively with something in the flagged file and was getting tired of copy and pasting from the JustNames csv export to flagged.txt.

This modified is_flagged function looks in the import_export folder (the created namepath variable points to that location) of the league save file and will pull the player IDs from the last modified or created csv file instead of looking at the flagged.txt.

Only other change I had to make was 'import os' for finding all of the csv's in the import_export folder.

def is_flagged(df: pd.DataFrame) -> pd.DataFrame:
    # Read player_ids from text file and convert to integers
    
    filenames = [file for file in os.listdir(namepath) if os.path.isfile(namepath / file)]
    for x in range(len(filenames)):
        filenames[x] = os.path.abspath(namepath / filenames[x])

    latest_file = max(filenames, key=os.path.getmtime)
    print(f"✅ Flagging File: ", latest_file)
    
    with open(latest_file, 'r') as f:
        flagged_ids = [int(line.strip()) for line in f if line.strip().isdigit()]
        f.close()

    # Add 'flag' column based on player_id match
    df["flag"] = np.where(df["player_id"].isin(flagged_ids), "flag", "")
    return df

[darkcloud4579]

Only finding this now, but appreciate all the work and effort you've put into it. Will take a look.