Sabitlenmiş Tweet
this might be the greatest thing i’ve posted so far on twitter
there’s actual data here + some real alpha that should help anyone farming on @extendedapp improve their PPM meaningfully
quick follow-up with real numbers from epochs 33–34
baseline first (so we’re anchored):
median across everyone ≈ 279 PPM
now the interesting part
volume doesn’t scale linearly with points
points follow a power law pretty cleanly:
Points ≈ 0.0010 × Volume^0.9009
R² ≈ 0.82
translation:
doubling your volume gets you less than 2× points
you can see the decay clearly by bucket:
• < $100k → ~360 PPM
• $100–500k → ~285
• $500k–1M → ~259
• $1–2.5M → ~254
• $2.5–5M → ~241
• $5–10M → ~232 (small sample)
• > $10M → ~220 (small sample)
personally, the $1–2.5M range lines up best with how i trade, it looks like a good scale without killing efficiency
taker ratio was a surprise
median PPM by taker mix:
• 0–20% taker → ~184
• 20–50% → ~215
• 50–80% → ~330 (best)
• 80–95% → ~287
• 95–100% → ~302
going full 100% taker isn’t optimal. not just from a cost angle, but also from a pure points efficiency standpoint
holding time (messier, but directionally clear)
this one’s noisier and needs more data, but pure churn clearly isn’t winning
median PPM by hold time:
• <15 min → ~273
• 15–30 min → ~253
• 30min–1h → ~261
• 1–2h → ~338
• 2–4h → ~306
• 4–8h → ~321
• 12–24h → ~391 (best)
• >24h → not enough data yet, but PPM was even higher
directionally: longer holds are spam trading, where the ceiling is… still TBD
positions is where things get spicy
this one stood out the most and honestly caught me off guard
median PPM by number of positions:
• 1–5 positions → ~587 PPM
• 6–15 → ~350
• 36–75 → ~210
that gap is massive. not sure yet if this is fully causal or a side-effect of something else, but it’s extremely consistent across the data
same story with pairs:
• 1 pair → ~315 PPM
• 2 pairs → ~280
• 6+ pairs → ~255
spraying across pairs looks strictly worse than focusing on one or two pairs
quick disclaimer
this isn’t clean science,data isn’t clean and i’m not pretending this explains everything but the patterns keep lining up the same way for the limmited amount of data i have
and obviously this doesn’t capture everything boosts, pnl, etc...
TL;DR
• points follow a power law with volume (diminishing returns)
• 50–80% taker beats full taker
• longer holds > pure churn
• fewer positions + fewer pairs = way higher PPM
i’ll be running this exact setup in the new epoch and sharing how it actually plays out
English
