# A2.28 The size lever, with the live exit (Analyst 2)

## Pre registered rule (written 18:30 Bangkok, 6 Oct, BEFORE the run)
- Live everything: wide gate, buy at gate + 30 s, 0.3x stop, non-kit sell all at 5x, Kit A and Kit B hold (50% trail, pull, 48 h), slot rule 3 + 2, exits at the end of the next block, real tick depth both ways, 1% fee each way.
- Tickets: $500, $1k, $2k, $3k, $5k.
- A size COUNTS only if $ a day is above $0 in BOTH halves AND the worst day is above minus 10x the ticket in BOTH halves.
- Wallet = 8x ticket (the go-live rule). Efficiency = $ a day divided by wallet.
- Answers: (1) among sizes that count, the ticket where $ a day per $ of wallet is highest; (2) the smallest ticket that counts AND makes $2,000 a day or more in BOTH halves.
- Fill loss vs paper: paper = same entry and exit moments at zero size (pool price only, fees in). Entry loss = tokens we get vs tokens at the pool price. Exit loss = $ we get vs tokens times the pool price. Own buy move = pool price after our buy vs before; count trades where it is more than 10%.

## Answer in short (18:31 Bangkok)
- All 5 sizes count ($ a day above $0 in both halves, worst day well inside minus 10x ticket).
- Smallest ticket that makes $2,000 a day in BOTH halves: **$1k** (+$3,317 / +$6,683). $500 misses half A (+$1,745).
- $ a day per $ of wallet is highest at the SMALLEST ticket ($500: 0.44 / 0.88 a day per $ of wallet) and falls with every step up. There is no hump: each extra dollar loses a little more to the pool. So "efficiency peak" is not the right way to choose size. Choose by how much drawdown you accept: max drawdown is about 4.2 to 4.7 tickets at every size.
- Each step up still adds a lot of $ a day: going from $1k to $2k adds +$2,685 / +$5,455; $2k to $3k adds +$2,151 / +$4,439; $3k to $5k adds +$1,502 / +$3,231 per extra $1k.
- The pool pushes back hard from $3k: our own buy moves the price more than 10% in 89% / 79% of trades at $3k, and 100% at $5k (0% at $2k). Live fills at $3k and up depend more on the pool depth at the moment of the buy.
- My pick for the ramp: $1k now, $2k after live fills match paper, $3k as the top unless live depth is better than this history. $5k gives more $ but gives away 61% / 89% of a ticket per trade to the pool.

## Size table ($ a day, live exit, slot rule, every friction)
| Ticket | $ a day A | $ a day B | per $ wallet A | per $ wallet B | worst day A | worst day B | max drawdown A | max drawdown B | counts |
|---|---|---|---|---|---|---|---|---|---|
| $500 | 1,745 | 3,511 | 0.44 | 0.88 | 0 | -1,574 | -2,124 | -2,348 | yes |
| $1k | 3,317 | 6,683 | 0.41 | 0.84 | 0 | -3,161 | -4,257 | -4,700 | yes |
| $2k | 6,002 | 12,138 | 0.38 | 0.76 | 0 | -6,367 | -8,553 | -9,412 | yes |
| $3k | 8,153 | 16,577 | 0.34 | 0.69 | -137 | -9,615 | -12,884 | -14,137 | yes |
| $5k | 11,156 | 23,039 | 0.28 | 0.58 | -2,271 | -16,224 | -21,642 | -23,619 | yes |

- Wallet: with an 8x ticket wallet no trade was ever skipped for cash; 5x was the least that never skipped at every size.
- $ a trade: see the csv. Half B still holds 2 open Kit B holds (A2.24), which lifts half B.

## What the pool takes (vs paper at zero size)
| Ticket | entry loss | exit loss (mean, A / B) | exit loss (median) | x lost vs paper A / B | own buy moves pool >10% A / B |
|---|---|---|---|---|---|
| $500 | 0.9% | 6.6% / 9.9% | 0.7% | 0.12 / 0.20 | 0% / 0% |
| $1k | 1.8% | 7.6% / 10.9% | 1.4% | 0.19 / 0.30 | 0% / 0% |
| $2k | 3.4% | 9.4% / 12.7% | 2.8% | 0.31 / 0.47 | 0% / 0% |
| $3k | 5.0% | 11.1% / 14.3% | 4.1% | 0.42 / 0.63 | 89% / 79% |
| $5k | 8.0% | 13.9% / 17.2% | 6.4% | 0.61 / 0.89 | 100% / 100% |

- The mean exit loss is high even at $500 because a pull leaves nothing to sell into (100% loss on those few trades). Without pulls the exit loss is 1.1% at $500, 2.2% at $1k, 4.2% at $2k, 5.9% at $3k, 9.0% at $5k.
- Big winners lose the most on the way out: a hold sold at 10x is a 10x bigger sell. At $5k a trailed hold loses 17.9% (median) on exit.
- The jump from 0% to 89% for "moves the pool more than 10%" between $2k and $3k is real: every gate is at the same size ($1M sellable), so pool depth at entry is nearly the same for every coin (median own move: 3.6% at $1k, 7.2% at $2k, 11.0% at $3k).

## Matches earlier work
- The $500 to $3k rows match A2.24 to the dollar (same trades, same sim). New here: $5k, wallet per size, and the fill-loss split.

## Limits
- 70 trades, 23 days. Live pools may be deeper or thinner than this history.
- Our own earlier sells and buys are not in the pool state of later trades (trades rarely overlap in the same coin).

## First principles score: 90%
Money per day = trades a day times edge per trade times size, minus what the pool takes. The pool's cut grows with size (entry about 1.7% per $1k), so $ per $ of wallet must fall as size rises. The table shows exactly that. The ramp choice is about drawdown and live depth, not a hidden peak.

Files: http://159.223.88.168/hx-c9ois/sw/sw-a2-28-size-table.csv

Analyst 2, 6 Oct 2026 18:31 Bangkok.