DRAFT_DAY_PLAN.md

🎯 Draft Day Plan — Pick #2

NSBA4 Draft Day Plan — Pick #2

Prepared 2026-06-06 for the 06-07 draft. You (xpoes) are the non-playing GM, drafting 2nd overall. Snake, 14 teams × 6 rounds. Embargoed until after the SSB draft.

TL;DR — the one play

Trade #2 to Connor for his #3 + a mid pick. Let Connor take Rohan. Take Akhil Batchu at #3.

The logic in one line: Rohan is priced correctly by the field (going top-2 = his true value), so sell him to a buyer; Akhil is the night's biggest mispricing (model #1, field #7–8), so capture him a slot later and pocket an extra pick.

Fallback: if Akhil is somehow gone at #3, take Harry G (same projected value, clean combine).

Your picks (snake slot 2): #2, 27, 30, 55, 58, 83. After the trade: #3, 27, 30, 55, 58, 83, + Connor's comp pick.

The top tier — four players inside the noise

All projected value, identity reliability, and last-season production:

These four are within sampling noise of each other on value (0.61–0.67). The edge is not in splitting them — it's in buying the one the field misprices.

Akhil case study — de-risked, and the night's biggest arbitrage

Anurag flagged "Akhil isn't top-3." We stress-tested it hard, and the model wins this one — here's the full chain:

• The worry: Akhil's value leaned on 6 games linked to him by a first-name-only identity match (confidence 0.60), and the model read him as ESS-dominant while his combine said physics (phys 8 was his top category). Looked like a possible phantom.
• Resolved by triangulation:
• PPG check: the "akhil" game record = #3 of 50 in nsba3 points-per-game (16.7). Elite, real production.
• His own self-scout: "t5 ppg last year, t3 reg season… a bit weaker on math/phys." He names himself an ESS/bio/chem main, weak at math/phys — which matches the model's ESS-dominant θ exactly. The combine phys-8 was the red herring; the game tape is the truth (D7: game > combine).
• No competing "Akhil" exists in nsba3, so the fuzzy match is almost certainly him.
• The mispricing: Anurag's board slots Akhil at chem #6 — a mid role player — because it doesn't credit his ESS. That's why Akhil himself predicts he falls to 7–8. Model #1, field #7–8.

Implication: do not spend the #2 pick on Akhil — that's paying sticker price for a marked-down player. Trade back toward his ADP, still land him, and bank the difference.

The trade — mechanics & price

Connor (pick #3) wants Rohan, who's going top-2 and won't reach him. That's your leverage.

• Structure: swap R1 picks (#2 ↔ #3); Connor adds compensation. Connor takes Rohan at #2; you pick #3.
• Pick-value math (curve is steeply convex — findings 22): dropping #2 → #3 costs 6.0 generic points. Connor's cheapest asset (#59, worth 11.9) more than doubles that. Your real cost is ~zero, because Akhil and Harry both survive to #3.
• Ask: #3 + Connor's #31 (R3, worth 20.2) as your opener; settle anywhere down to #3 + #59. Any added pick is profit.
• Why it's clean: you sell Rohan at full market value (the field rates him top-2) and keep Akhil at a discount. Selling the fairly-priced asset to buy the underpriced one is the whole game.

Risk: only bites if Gideon snipes Akhil at #1 and Connor takes Harry instead of Rohan — and Connor's whole premise is wanting Rohan. Low. (And if Gideon takes Akhil, you'd fall to Harry G at #2 or #3 — the trade doesn't worsen that branch.)

Field model — reading Anurag's board for arbitrage

Anurag's subject-tier board is a window into how a sharp slice of the field drafts. Cross-referenced with our value model:

• Field under-rates (model-high → wait / trade back): Akhil (the big one — Anurag chem #6 vs model #1).
• Field and model agree (will go early — take early if you want them): Ishaan Kabra (Anurag ESS #3 / model #5), Kian.
• Anurag higher than model — and scibowl backs Anurag: Theenash, Roshan (see below). These may not slide as far as our board suggests.
• On Anurag's board but not (yet) draftable — verify eligibility before planning around them: Rahib H, William W, Lishan P, Sanjay O (unresolved or unpaid in our pool).

MSJ buy-lows — corroborated, but temper the slide

The scibowl data (league avg celerity 0.088) shows two MSJ players producing far above their combine:

Our model undervalues them (it uses scibowl only for difficulty priors, not per-player ability). But Anurag independently rates them high (Theenash bio #4 and chem #5 — a generalist; Roshan ESS #1). So the field is not asleep on them — they likely won't fall to rounds 4–5. If you want them, plan to reach in R2–R3 (#27/#30), not later. Ishaan Kabra (#5, 0.507) is the model-endorsed MSJ option if you want one earlier.

Mid-round plan (R2–R3) — cover Akhil's holes

Akhil is ESS/bio/chem and weak math/phys by his own admission. Filling that is coverage logic, not synergy (duo "chemistry" washed out — findings 29; you draft the best player who fills a hole, not a mystical pair).

• Best available math/phys fills: Praneel Avula (phys θ 5.0, val 0.368 — gettable R3–4), Vishnu M (math θ 2.3, val 0.399), Chris Wang (phys 2.8 / math 1.3).
• ⚠️ Akhil suggested Edward Chen (Carmel) as his math/phys pairing — the data disagrees. Edward grades weak (val 0.235; math θ 0.4, phys θ 0.1) and isn't in the draftable pool. Trust Praneel/Vishnu over the player's roommate rec.

Caveats — trust directions, not decimals

• Availability is the biggest unmodeled risk. Across 216 past drafted player-seasons: 39% played zero games, median attendance 30%, and it is not predictable (season-to-season r ≈ −0.2). No model fixes this — lean on commitment/Discord signals. A point in Akhil's and Rohan's favor: both are returners who've shown up before.
• The combine "swing": the −1 itself is mostly noise (r ≈ 0.10–0.15 with game outcomes); the +4 superpower rate is the real signal (r = 0.34 with game value) and is already partly in the model. Aggression-as-negs ≠ much; aggression-that-converts = the thing.
• Small samples everywhere: 6-game tapes, 41 team-seasons. Identity merges include some fuzzy links (Akhil's was 0.60 — here corroborated, but not all will be).
• Combine still closing. Re-pull the board once final before you're on the clock.

Live tooling

.venv/bin/python scripts/draft_live.py recommend        # advice for your next pick
.venv/bin/python scripts/draft_live.py board             # full pool, your marginal value
.venv/bin/python scripts/draft_live.py board --subject phys
.venv/bin/python scripts/draft_live.py pick "Kian Dhawan"   # log any pick
.venv/bin/python scripts/draft_live.py me "Akhil Batchu"    # log YOUR pick
.venv/bin/python scripts/draft_live.py undo

Paste picks as they happen and I'll run it live. By R2 (#27) the engine shifts from best-value to coverage-marginal weighting automatically.

NSBA Draft Analytics · embargoed until after the SSB draft · ← hub