Week 2 was mean reversion; Week 3 was a regime break. ChessFun, czz, and sumin โ three of the four richest books on the Street โ all lost, two of them by sixty-plus. This close we're marking every desk against its benchmark โ the expected P&L and winning tickets a book should have booked given exactly who sat the desk and how strong each asset class was. It turns the blowups into an attribution: some shops smashed their number, others cratered miles below it. The two factors that keep driving returns in this league were the same two again โ who showed up on the desk and who kept the risk limits.
For every desk we simulate the winning tickets it should have booked given the exact PMs who dressed and each side's asset-class strength, then mark it against what it actually printed. Positive = alpha (booked more tickets than the matchup implied); negative = negative alpha. It's the cleanest "was it real?" filter on a wild close.
| Desk | Actual tickets | Expected (bench) | Alpha (ฮ) | The mark |
|---|---|---|---|---|
| Connor Chang | 16 | 9.9 | +6.1 | Kian D's 11-for-11 dragged the whole book miles past its number. Utterly real alpha. |
| cryo | 12 | 9.1 | +2.9 | The $60M limit-down on czz was earned โ cryo beat its benchmark across every desk. |
| dan.k.memes | 12 | 9.6 | +2.4 | Won the volume battle by more than the matchup implied. 3โ0 is no fluke. |
| anurag | 11 | 9.0 | +2.0 | Finally monetized the talent โ genuine alpha, not just sumin fading. |
| jonathan | 10 | 8.6 | +1.4 | The upset had substance โ out-booked its benchmark even with 7 blow-ups. |
| xpoes | 9 | 7.8 | +1.2 | A PM short and STILL beat its number. Imagine this book with a full desk. |
| GidTheKid2 | 10 | 9.6 | +0.4 | Right on benchmark; the win came from an 80% capture night, not extra tickets. |
| Fez_Keyreb | 7 | 7.7 | โ0.7 | Roughly par on coverage; the loss was blow-ups and ticket volume, not sourcing. |
| ChessFun | 8 | 8.9 | โ0.9 | Slightly under par โ and that's WITH Eric L off the desk lowering the bar. Headcount cost them. |
| sumin | 7 | 8.8 | โ1.8 | Negative alpha and lost the unbeaten tag. anurag's Credit sweep did real damage. |
| czz | 6 | 9.0 | โ3.0 | The $60M loss wasn't variance โ czz printed 3 tickets below its own floor. |
| (Stitch)ยฒ | 4 | 8.2 | โ4.2 | The โ$14M was a total-system failure: 4 tickets on an 8-ticket desk, plus ten blow-ups. |
The tape of the season, in both directions. Kian D went 11-for-11 for $44M with zero blow-ups โ the best single session anyone has printed โ and he didn't do it in a vacuum: Connor swept Commodities 4โ0 (against a benchmark of just 2.0 there) and took Equities, Credit and Derivatives 3โ1. As a book Connor printed 16 winning tickets against an expected 9.9, a +6.1 alpha run that is the largest of the week and removes any doubt this was a scheduling fluke. This is monetization on top of sourcing on top of a star PM having the session of his life.
(Stitch)ยฒ was the mirror image โ a total-system failure and a book that finished underwater. They managed just 4 tickets on a desk worth ~8 (โ4.2), then blew up ten times for a โ$40M self-inflicted drawdown that turned a bad close into a historic one: final mark โ$14M. pancreas uvula alone accounted for 5 of the ten blow-ups. When you hand back forty million on interrupts, the asset-class battle is almost beside the point โ the risk desk lost the session before the sourcing desk ever got a vote. Lean in: this is what a negative book looks like.
The West statement of the close, and the alpha says it was fully deserved: cryo booked 12 vs 9.1 expected (+2.9) while czz cratered to 6 vs 9.0 (โ3.0) โ a six-ticket swing that is the entire $60M margin. The engine room was Credit, where cryo went 4โ0 against a 1.8 benchmark (+2.2), the single most dominant asset-class print of the round. When one desk owns a whole book of the tape, the margin takes care of itself.
cryo also owned the clock, executing at an average of word 20 to czz's 34 โ lower latency to the bid on nearly every look. czz never found a second PM behind vishnu m (4 tickets); the supporting desk printed just $20M combined. A title contender looked ordinary, and the model agrees it wasn't bad luck โ it was negative alpha on a one-PM book.
The upset of the week is really a headcount story. ChessFun's Eric L โ its lowest-latency, most productive PM โ booked zero tickets (off the desk), and the top-heavy book never recovered. Even with the bar already lowered by his absence, ChessFun still marked below par (8 vs 8.9, โ0.9) โ there was no next man up on the bench.
jonathan pounced, and it wasn't a smash-and-grab: they ran alpha at 10 vs 8.6 (+1.4), powered by a Credit 3โ1 edge (3 vs 1.5 expected) and a Derivatives 2โ0 sweep. The wart is risk discipline โ seven blow-ups โ the only reason this was $18M and not $30M. Clean up the interrupts and the Week-2 giant-killer is a genuine problem; a 60% capture night to ChessFun's 38% shows the sourcing is there.
sumin's unbeaten aura ended hard, and the alpha says it was more anurag rising than sumin collapsing โ anurag ran +2.0 (11 vs 9.0) while sumin came in modestly under (7 vs 8.8, โ1.8). The decisive book was Credit: anurag went 4โ0 against a 1.6 benchmark (+2.4), the biggest single-desk overperformance of the session. This is the anurag book finally monetizing the talent it out-sourced with in Week 1.
sumin's one bright spot was a Derivatives sweep (3โ0, +1.8 over bench), but they got shut out in Commodities (0 vs 1.6) and never threatened anurag's coverage anywhere else. Sanjay O (4 tickets, low latency at word 19) ran the anurag desk. After an 0โ2 open built on fixable back-office mistakes, anurag turned the corner and took the East's last unbeaten book down with it.
The most interesting alpha line of the week: both books ran positive. Gid landed right on its number on tickets (10 vs 9.6) and won on the capture line instead โ an elite 80% monetization (8/10) โ plus Credit and Derivatives 3โ1 edges. This was a win built on cashing what they sourced, not on extra volume.
The story underneath is xpoes: 9 tickets against a 7.8 benchmark (+1.2) while a PM short (no Rohan G, again). Jianyu was the session's second-best PM (5 tickets) and they swept Equities 3โ0 (+2.1 over bench) โ a genuinely strong print from a thin book. Whole, this xpoes desk beats its number by more and maybe wins the session; short, it ran out of coverage in Rates (0 vs 1.3) and couldn't match Gid's back office.
The cleanest "capture isn't everything" session of the year. Fez_Keyreb cashed 71% of bonuses to dan's 42% โ and lost by $36M, because dan simply booked far more tickets: 12 vs 9.6 expected (+2.4) to Fez's par-ish 7 vs 7.7 (โ0.7). You can't monetize deals you never source.
dan did it with sweeps โ Systematic 3โ0 and Commodities 4โ0 (+2.3 over bench there) โ and, crucially, a clean sheet on risk limits (0 blow-ups) while Fez gave four back. varnite (5 tickets) paced a balanced book with no weak desk. dan.k.memes is a spotless 3โ0, and the benchmark calls it the most consistent shop in the league โ it beats its number every session without ever looking spectacular. The disciplined compounder.
All 14 books marked to market after Week 3, both desks. Records first, but the marks lean on alpha โ who's really as good as their P&L.
EAST
| # | Book | The Mark |
|---|---|---|
| 1 | dan.k.memes | 3โ0 and beats its benchmark every session (+2.4). Wins on volume and sweeps, not capture (42%). The standard. |
| 2 | GidTheKid2 | 2โ1 with Harry Gao back on the desk and an 80% capture night. Right on benchmark; the ceiling is the draft-day book. |
| 3 | mingle/Yunyi | Dark this week, 1โ1, still owns the $130M single-session high. Untested since the capture fix โ reopens next week. |
| 4 | sumin | 1โ1. Ran negative alpha (โ1.8) losing the unbeaten tag. Still holds that $138M, but the Credit hole is real. |
| 5 | anurag | 1โ2, but the eye test and the alpha agree the book is rising (+2.0). A $30M win that ended sumin's run. |
| 6 | xpoes | 1โ2, and the most unlucky book on the Street โ beats its number (+1.2) every session while a PM short. |
| 7 | Fez_Keyreb | 0โ2. Par on coverage, but blow-ups and low ticket volume keep sinking a strong capture stroke. |
WEST
| # | Book | The Mark |
|---|---|---|
| 1 | cryo | 2โ1 and the hottest book on the Street โ +2.9 alpha, a $60M limit-down of a contender at blazing latency. |
| 2 | ChessFun | 2โ1, beat cryo head-to-head โ but got upset and marked below par with Eric L off the desk. The headcount watch is everything. |
| 3 | czz | 2โ1, but โ3.0 in a $60M loss โ not variance, a genuine no-bid session behind vishnu m. Too much single-PM concentration. |
| 4 | jonathan | 1โ1 and felled the giant with substance (+1.4). Fix the 7 blow-ups and this spoiler book is dangerous. |
| 5 | James W | Dark this week, 1โ1. The preseason A+ still finding its level; a real test at ChessFun awaits. |
| 6 | Connor Chang | 1โ2, but +6.1 alpha and Kian D's $44M say the record is a lie. Both losses were to the West's best books. |
| 7 | (Stitch)ยฒ | 0โ2 off a โ$14M (โ4.2 alpha, ten blow-ups). The talent isn't the problem; the risk desk is a five-alarm fire. |
The chalk went 5โ1 in Week 2; Week 3 shredded it again. Three favored contenders โ czz, sumin, and ChessFun โ all lost, and the alpha ledger backs the carnage: czz (โ3.0) and sumin (โ1.8) genuinely ran negative, while the shorts that beat them (cryo +2.9, anurag +2.0, jonathan +1.4) all cleared their benchmark. The tape that faded the crowded long paid: cryo (a home dog by our number) won by $60M, and jonathan flipped an 11-point deficit into an $18M win. Reputation is worth less here than headcount and risk discipline on the night โ and now we can mark exactly how much less. The Street never repriced ahead of the print; odds barely drifted toward the dogs intraday and only gapped once the P&L locked.
| PM | Desk | P&L | Tickets | Blow-ups | Avg latency |
|---|---|---|---|---|---|
| Kian D | Connor Chang | $44M | 11 | 0 | 21 |
| Jianyu | xpoes | $20M | 5 | 1 | 20 |
| varnite | dan.k.memes | $20M | 5 | 0 | 20 |
| Sanjay O | anurag | $16M | 4 | 0 | 19 |
| suzuko | cryo | $16M | 4 | 1 | 16 |
| Harry Gao | GidTheKid2 | $16M | 4 | 1 | 31 |
โก Lowest latency: suzuko (word 16) and anurag's Sanjay O (19) set the pace to the bid. Risk-limit trouble: pancreas uvula ((Stitch)ยฒ, 5 blow-ups) headlined the ten-blow-up implosion; Abhinav (Gid) had 3 and Chris W (Fez) had 2 in a losing seat.
Week 4 puts the two hottest storylines on the marquee and sits Connor Chang and anurag dark. The questions into the open: can anyone catch dan.k.memes (3โ0) in the East, and is Eric L back on the ChessFun desk?
The East's unbeaten book meets its scariest ceiling. dan.k.memes (3โ0) wins the quiet way โ volume, sweeps, zero blow-ups, +2.4 alpha โ while mingle/Yunyi, reopening after a dark week, owns the league's single-session high ($130M) and a fixed capture stroke. dan's relentless coverage against mingle's top-end firepower; win it and you're the clear class of the conference.
ChessFun's ceiling rides on headcount. With Eric L back on the desk this is a top-two West book; without him, jonathan just published the blueprint and the alpha (โ0.9 even at a lowered bar) says there's no next man up. A rested, diversified James W is exactly the kind of book that punishes a short-handed favorite.
Two redemption tapes. The hottest book on the Street, cryo, visits (Stitch)ยฒ โ can a โ4.2, ten-blow-up desk stabilize its risk in a week? And czz, freshly humbled by $60M and 3 tickets below its floor, draws the giant-killer jonathan โ a nasty landing spot to prove the blowup was a blip, not a level-set.
Both sit out Week 4. Connor's +6.1 alpha and Kian D's $44M mean the 1โ2 record badly understates the book โ this is a top-half desk marking time. anurag exits on a rising note (+2.0, ended sumin's unbeaten run); the momentum has to survive a week in cold storage.