How to compute expectancy in a dataframe across rows

I have a dataframe that contains day, symbol, strategy, and pnl. I want to analyze and compare pnl in a couple of ways.

I'd like to get the win-rate & expectancy when grouped by symbol and strategy. So I've done this:

def stats(s):
    winrate = s['isWinner']['count'] / (s['isWinner']['count'] + s['isLoser']['count'])
    expectancy = s['isWinner']['mean'] * winrate - s['isLoser']['mean'] * (1.0 - winrate)

df["isWinner"] = df['pnl'] >= 0
df["isLoser"] = df['pnl'] < 0
df2 = df.groupby(['day', 'symbol', 'strategy', 'isWinner']).agg({'pnl': ['count', 'mean', 'std', 'min', 'max']})
df2.groupby(['day', 'symbol', 'strategy']).agg(stats)

Apparently, I can't do s['isWinner'] in the stats function. What am I doing wrong?

Once the stats function works, how do I add winrate and expectancy to df2?

Am I going about this the right way? Is it necessary to create df2 from df, or is there a better way?

I'm sure there is a more pythonic way to do this, but it works.

    df["isWinner"] = df['pnl'] >= 0
    df["isLoser"] = df['pnl'] < 0

    grouped = df.groupby(['strategy'])
    stats = pd.DataFrame()
    for name, grp in grouped:
      wincount = grp['isWinner'].values.sum()
      loscount = grp['isLoser'].values.sum()
      winrate = wincount / (wincount+loscount)
      profits = grp[grp['isWinner']]['pnl'].sum()
      losses = grp[grp['isLoser']]['pnl'].sum()
      avgwin = profits/wincount
      avglos = losses/loscount
      expectancy = winrate * avgwin + (1.0 - winrate) * avglos
      row = {'name': name, 'wincount': wincount, 'losscount': loscount, 'winrate': winrate, 'profits': profits, 'losses': losses, 'avgwin': avgwin, 'avgloss': avglos, 'expectancy': expectancy}
      stats = stats.append(row, ignore_index=True)

    stats

Results: