def whatif(lb, users, s, c):
"""
Profit if `users` had staked `s` and `c` in every tournament.
Earnings are left in NMR instead of splitting the NMR earnings into
NMR and USD.
"""
if isinstance(users, list):
pass
elif nx.isstring(users):
users = [users]
else:
raise ValueError("`users` must be str or list (of str)")
cols = ['nmr_staked', 'nmr_burn', 'nmr_earn', 'nmr_net']
df = pd.DataFrame(columns=cols)
lb.insert(0, 'pass', lb['live'] < LOGLOSS_BENCHMARK)
rounds = np.sort(lb['round'].unique())
for r in rounds:
d = lb[lb['round'] == r]
if r > 112:
staked = 0
burn = 0
earn = 0
for t in nx.tournament_all(as_str=False):
dt = d[d.tournament == t]
if dt.shape[0] > 0:
cutoff, ignore = calc_cutoff(dt)
if c >= cutoff:
idx = dt.user.isin(users)
dti = dt[idx]
idx = dti['pass']
nwin = idx.sum()
nlos = (~idx & (dti['live'].notna())).sum()
p = (1.0 - cutoff) / cutoff
burn += nlos * s
earn += nwin * s * p
staked += idx.size * s
net = earn - burn
df.loc[r] = [staked, burn, earn, net]
else:
raise ValueError("`round1` must start at at least 113")
df.loc['total'] = df.sum()
return df
def cutoff(lb):
"Independent calculation of confidence cutoff"
cols = nx.tournament_all(as_str=True)
df = pd.DataFrame(columns=cols)
rounds = np.sort(lb['round'].unique())
for r in rounds:
d = lb[lb['round'] == r]
if r > 112:
cut = []
for t in nx.tournament_all(as_str=False):
dt = d[d.tournament == t]
cutoff, ignore = calc_cutoff(dt)
cut.append(cutoff)
else:
cut = [np.nan] * 5
df.loc[r] = cut
df['mean'] = df.mean(axis=1)
df.loc['mean'] = df.mean()
return df
def pass_rate(lb):
"Fraction of users who beat benchmark in each round"
cols = ['all', 'stakers', 'nonstakers', 'above_cutoff', 'below_cutoff']
df = pd.DataFrame(columns=cols)
rounds = np.sort(lb['round'].unique())
for r in rounds:
d = lb[(lb['round'] == r) & (lb.live.notna())]
d.insert(0, 'pass', d['live'] < LOGLOSS_BENCHMARK)
pr_all = d['pass'].mean()
pr_stakers = d[d['s'] > 0]['pass'].mean()
pr_nonstakers = d[d['s'] == 0]['pass'].mean()
if r > 112:
nabove = 0
nbelow = 0
pabove = 0
pbelow = 0
for t in nx.tournament_all(as_str=False):
dt = d[d.tournament == t]
cutoff, ignore = calc_cutoff(dt)
nabove += dt[dt.c > cutoff].shape[0]
nbelow += dt[dt.c < cutoff].shape[0]
pabove += dt[(dt.c > cutoff) & (dt['pass'])].shape[0]
pbelow += dt[(dt.c < cutoff) & (dt['pass'])].shape[0]
if nabove == 0:
pr_above = np.nan
else:
pr_above = 1.0 * pabove / nabove
if nbelow == 0:
pr_below = np.nan
else:
pr_below = 1.0 * pbelow / nbelow
else:
pr_above = np.nan
pr_below = np.nan
df.loc[r] = [pr_all, pr_stakers, pr_nonstakers, pr_above, pr_below]
df.loc['mean'] = df.mean()
return df
def payout(lb):
"NMR and USD payouts per round"
cols = [
'staked_nmr', 'staked_above_cutoff', 'burned_nmr', 'nmr_payout',
'usd_payout', 'total_payout_in_nmr'
]
df = pd.DataFrame(columns=cols)
rounds = np.sort(lb['round'].unique())
lb.insert(0, 'pass', lb['live'] < LOGLOSS_BENCHMARK)
for r in rounds:
d = lb[lb['round'] == r]
if r > 112:
nmr_cut = 0
nmr_cut_pass = 0
for t in nx.tournament_all(as_str=False):
dt = d[d.tournament == t]
cutoff, ignore = calc_cutoff(dt)
nmr_cut += dt[dt.c >= cutoff].sum()['s']
nmr_cut_pass += dt[(dt.c >= cutoff) & (dt['pass'])].sum()['s']
else:
nmr_cut = np.nan
if cutoff == 0:
total = np.nan
else:
total = nmr_cut_pass * (1.0 - cutoff) / cutoff
ds = d.sum()
pay = [
ds['s'], nmr_cut, ds['nmr_burn'], ds['nmr_stake'], ds['usd_stake'],
total
]
df.loc[r] = pay
fraction = df['burned_nmr'] / df['staked_above_cutoff']
df.insert(3, 'fraction_burned', fraction)
df.loc['mean'] = df.mean()
df = df.round(2)
return df