34 — NSBA1 (2022) Draft Analysis & Win-Shares Board
34 — NSBA1 (2022) Draft Analysis & Win-Shares Board
Season: nsba1_2022 · Source: full competition (actual game results, not combine)
Outputs: outputs/nsba1_winshares.csv
Data caveats (read first)
- NSBA1 game sheets use NICKNAMES only — no Discord tags, no real names. Win-shares
are keyed by in-game nickname. Most NSBA1 game players are singletons in
canonical_players.csv. - Weak draft→game link. Of 72 NSBA1 draft picks, only 12 (16.4%) link to a realized win-shares record (consistent with the project-wide ~19% clean-map rate / median roster Jaccard ~0.11). The other 61 picks cannot be linked to a game record.
- 5 scoring categories in NSBA1: Bio (b), Chem (ch), Energy (ene — NOT CS), Earth/Space (ess), Math (m), Physics (p). CS did not exist this season.
- Win-shares are RAW (toss-up production based) and subject to a usage/empty-stats
caveat (lone stars on weak teams can be over-credited). A teammate-adjusted version is
being computed separately (
usage_bias_32.csv/adjusted_ability.csv). Notes on that below where relevant.
Win-share = point_share × team_wins, where point_share is a player's share of their
team's toss-up points.
1. NSBA1 Win-Shares Board (top 25 of 78 game players)
| Rank | Nickname | Team | G | PPTF | Win Shares | Team Win% |
|---|---|---|---|---|---|---|
| 1 | arolakiv | devansh | 10 | 1.207 | 5.10 | .636 |
| 2 | VulcanForge | sheep b | 11 | 0.896 | 4.67 | .636 |
| 3 | Vish | mount ellis | 11 | 0.744 | 4.24 | .727 |
| 4 | AdiyaSivahuma | aastha | 11 | 0.774 | 3.62 | .455 |
| 5 | Sean29 | kabirs kaolinites | 10 | 0.727 | 3.37 | .636 |
| 6 | Anurag Sodhi | mount ellis | 11 | 0.562 | 3.20 | .727 |
| 7 | reverse | sample text | 9 | 0.681 | 2.69 | .727 |
| 8 | ne | doge disciples | 10 | 0.887 | 2.25 | .364 |
| 9 | Daniel S | Bellevue Biphosphoglycerates | 11 | 0.496 | 2.14 | .545 |
| 10 | Nipun | kabirs kaolinites | 9 | 0.442 | 1.86 | .636 |
| 11 | minitarrasque | sample text | 10 | 0.360 | 1.68 | .727 |
| 12 | Tiberius | sample text | 8 | 0.734 | 1.68 | .727 |
| 13 | Richard Zhu | Bellevue Biphosphoglycerates | 10 | 0.497 | 1.64 | .545 |
| 14 | TechnologicalError | okehios | 10 | 0.461 | 1.47 | .364 |
| 15 | Vishnu M | Bellevue Biphosphoglycerates | 11 | 0.335 | 1.43 | .545 |
| 16 | somath | sheep b | 11 | 0.296 | 1.38 | .636 |
| 17 | 4435 | aastha | 8 | 0.386 | 1.31 | .455 |
| 18 | Fez_Keyreb | ct610 | 10 | 0.527 | 1.28 | .273 |
| 19 | abcisosm5 | sample text | 10 | 0.286 | 1.26 | .727 |
| 20 | ANdrey | okehios | 10 | 0.382 | 1.24 | .364 |
| 21 | cryo | devansh | 10 | 0.288 | 1.22 | .636 |
| 22 | Fumarole | kabirs kaolinites | 11 | 0.214 | 1.10 | .636 |
| 23 | Fishface | con1596 | 10 | 0.490 | 1.00 | .273 |
| 24 | Gideon Tzafriri | con1596 | 7 | 0.594 | 0.92 | .273 |
| 25 | Fishface | nearest airport | 10 | 0.490 | 0.80 | .273 |
Full 78-player board (with per-category PPTF) in outputs/nsba1_winshares.csv.
2. Draft steals & busts (LINKABLE picks only)
Link rate: 12 of 72 picks (16.4%) link to a win-shares record. 61 picks could not be linked — for the unlinkable majority, draft value vs. realized production is simply unknowable from this data. Treat the lists below as suggestive, not comprehensive.
Among the 12 linkable picks, Spearman corr(overall_pick, win_shares) = −0.68 — later picks did genuinely produce less, so the market wasn't random, but the misses are large.
Biggest STEALS (late pick, high realized WS)
| Overall Pick | Rd | Team | Pick name | WS | Notes |
|---|---|---|---|---|---|
| 33 | 3 | mount ellis | Vish | 4.24 | #1 WS of all linkable picks, taken at pick 33 |
| 13 | 2 | Bellevue Biphosphoglycerates | Richard Zhu | 1.64 | solid round-2 value |
| 52 | 5 | sample text | Abhi | 0.59 | 5th-round contributor |
Vish is the headline steal: drafted 33rd overall (round 3), he produced the highest win-shares of any linkable pick in the draft, and ranked 3rd among all 78 game players.
Biggest BUSTS (early pick, low realized WS)
The linkable early picks (pick ≤ 15) actually all produced positively:
| Overall Pick | Rd | Team | Pick name | WS |
|---|---|---|---|---|
| 13 | 2 | Bellevue Biphosphoglycerates | Richard Zhu | 1.64 |
| 1 | 1 | doge disciples | ne | 2.25 |
| 9 | 1 | mount ellis | Anurag Sodhi | 3.20 |
No clean "early-pick, zero-WS" bust exists among the linkable subset — the true busts are almost certainly hiding in the 61 unlinkable picks (early picks whose drafted player never produced a matchable game record). The honest read: we can confirm steals but cannot confirm busts, because the players who busted are exactly the ones who don't link.
3. Vishwa Reddy Akkati — search result
Most likely match: "Vish" (canonical_id P0618), Mount Ellis — but needs human confirmation.
No record anywhere in the dataset contains the strings Vishwa, Reddy, or Akkati
(searched all of data/processed/). NSBA1 has no real names attached to game/draft records,
so a definitive string match is impossible. Candidate evaluation:
| Candidate | canonical_id | Why / why not |
|---|---|---|
| Vish | P0618 | Best candidate. "Vish" is the natural short form of "Vishwa." NSBA1-only (no later season under this id), which fits a player who never revealed a real name. Appears in both the NSBA1 combine and game sheets, and was drafted ("Vish", pick 33). No discord tag to confirm against. |
| Vishnu M / Vishnu Mangipudi | P0619 | Ruled out. Real name revealed in NSBA2/3 as Vishnu Mangipudi (discord oof7373) — not Akkati. |
| Vishal Surya | P0355 | Ruled out. Real name Vishal Surya (discord lavar ball#2773), NSBA2+. Not Akkati. |
| Reddy names | P0037, P0288, P0332, P0515 | Ruled out. All are SciBowl (ssb_2026) players, not NSBA1. |
If P0618 "Vish" is Vishwa Reddy Akkati, his NSBA1 profile is:
- Win shares: 4.235 → rank 3 of 78 game players (behind only arolakiv and VulcanForge).
- PPTF: 0.744 (toss-points per toss-up faced), 49 correct / 4 negs over 11 games, 242 TUH.
- Team: Mount Ellis, 72.7% win rate (8–3) — a strong, balanced team (teammate Anurag Sodhi was rank 6).
- Draft: taken 33rd overall, round 3 by GM karthik → biggest draft steal of NSBA1.
- Combine: overall score 6.0, eligible. Z-profile: Phys +5, Math +3, Bio +1, Chem −1, Earth/Space −2.
- Per-category in-game PPTF: Chem 1.27 (best), Energy 1.09, Earth/Space 0.82, Bio 0.64, Math 0.27, Physics 0.27. In-game he scored on Chem/Energy despite a physics-heavy combine — a generalist who leaned chemistry in live play.
- Robustness: under the teammate-adjusted view (
usage_bias_32.csv), adjusted PPTF = 0.705 vs raw 0.744, rank shift only −2 (obs rank 9 → adj rank 11). He is not an empty-stats artifact — Mount Ellis had a real second option (Anurag), so his production is credibly his own.adjusted_ability= 1.376 (reliable=True).
Confidence: moderate-to-high on the nickname inference, but NOT certain. "Vish" → "Vishwa" is plausible and there are no competing un-ruled-out candidates, but with zero real-name or discord linkage in NSBA1, a human who knows the league should confirm that the NSBA1 nickname "Vish" was Vishwa Reddy Akkati before this is treated as fact.
4. Other recognizable names worth surfacing
- Anurag Sodhi (P0093, Mount Ellis) — rank 6, 3.20 WS; Vish's teammate. Drafted 9th overall, lived up to it.
- Richard Zhu (Bellevue) — rank 13, full real name in NSBA1, drafted 13th, solid value.
- Vishnu Mangipudi ("Vishnu M", P0619) — rank 15; recurs across NSBA1–3, a useful cross-season anchor (distinct from Vish).
- Ashmit Dutta (P drafted to devansh, pick 39) — recognizable full name but only 0.08 WS.
- arolakiv (rank 1, 5.10 WS, PPTF 1.21) and VulcanForge (rank 2) — the two players ahead of Vish; both nickname-only, no draft link, identities unconfirmed.