Visual ranking of the current state by return since start — not a scientific winner. Bots that have not traded yet are listed last (–). SMALL SAMPLE = fewer than 30 closed trades. PnL = equity − start.
Active bots
Recent operations
Time
Bot
Coin
Action
Size
PnL / Result
Reason
Loading…
Tap a row for signal / quote MC, impact, fees, latency, trigger and links. Smart control books are not listed here (see Tournament details).
PAPER TRADING — real-time market data and executable quotes, no real orders.
LIVE MARKET / PAPER EXECUTIONexperiment —rules —started —
Waiting for the paper tournament snapshot...
Five frozen scenarios read the same live KOL wallet trades as LIVE NOW (PumpDev). Every entry and every exit goes through
signal → execution delay → fresh Jupiter quote for the exact size → paper fill: the fill is Jupiter's outAmount
minus an adverse-execution buffer, plus explicit network/priority/aggregator fees calibrated on real CoinLaunchr swaps.
If the quote fails (rate limit, timeout, no route, excessive impact) the signal is recorded as UNEXECUTABLE — no fill is invented.
Balances, equity and ranking use executable net PnL only; market PnL is shown as a diagnostic.
Experiment paper_v2_7_f3_variants_2026_10_05, pre-registered before its first trade
(doc/analisis/2026-10-05_paper_v2_7_f3_variants_prereg.md). Five entry rules on the same real Opportunity Alerts; same execution model,
same exit (F3's, verbatim), fixed $20 per entry and no bankroll cutoff. V0 = historical F3 rule, exact. Trades the arms share use one shared Jupiter
quote, so their returns are identical: arms differ only by which alerts they pick (see COMMON / VARIANT ONLY / CONTROL ONLY).
No verdict before 30 closed trades per arm; the rules never change between milestones.
Waiting for data...
Active paper scenarios
PAPER TRADING — NO REAL ORDERS. Every paper scenario that is still running, in a fixed logical order (never ranked by performance).
Status: ACTIVE = can trade and entered in the last 48 h · ACTIVE_NO_SIGNALS = can trade, no entry in 48 h · BANKROLL_EXHAUSTED = can never trade again ·
STOPPED = service not writing · ARCHIVED = retired on purpose (listed below under HISTORICAL / ARCHIVED).
Each experiment has its own $100 bankroll per scenario and its own start date, so PnL is never added across experiments. SIN MUESTRA = no closed trade yet (no win rate, profit factor or expectancy);
MUESTRA PEQUEÑA = fewer than 30 closed trades (figures are descriptive only).
Scenario
Family
Status
Trades
Open
Wins
Win rate
PnL
Balance
Profit factor
Expectancy
Max DD
Fees
Started
Loading...
HISTORICAL / ARCHIVED
Experiments retired on purpose. Their databases are kept unchanged and these are their final frozen figures;
they open no new positions. State at archive = what the scenario could still do when it was stopped. Exec. drag is computed over all closed trades.
Scenario
Family
Status
State at archive
Trades
Wins
PnL
Balance
Profit factor
Max DD
Fees
Exec. drag
Archived
Loading...
Executable Equity Curve
Paper balance per scenario: cash from executed fills plus open positions marked at an executable estimate (sell buffer + fees). Toggle "cash" to see realized only.
Waiting for data...
A–E · KOL WALLET TOURNAMENT
Experiment paper_v2_2026_09_25 only (A-CONSENSUS, B-CONVICTION, C-PULLBACK, D-MOMENTUM, E-SINGLE-KOL): frozen rules, $100 bankroll per scenario.
Latency is shown separated: BUY = trigger event → fresh Jupiter quote; SELL = the attempt that filled; sell retries and blocked time apart (a retried exit never turns into "BUY latency"). Descriptive only.
Waiting for data...
B-CONVICTION EXIT EXPERIMENTS
B (control) and B1–B4: same entry, bankroll, sizing and execution model; only the exit differs (executable partial exit). Descriptive only.
Waiting for data...
OPPORTUNITY ALERT ENTRY EXPERIMENTS
F (control) buys every real Opportunity Alert; F1–F3 add only an entry filter on fields persisted at alert time; G/H are selective alert entries. Same bankroll, sizing, execution model and exit policy as F. Descriptive only.
Waiting for data...
FORWARD ALERT FILTER EXPERIMENT
F0 — FORWARD CONTROL buys every alert F would buy, but only from its own start and with a fresh $100 bankroll; F1–F3 add only an entry filter.
Everything here uses only the common window (alerts emitted at or after F0 started), so bankroll, sizing, execution model, exits and time window are the same. F (legacy) is excluded on purpose: it started earlier with a depleted bankroll. Descriptive only.
Waiting for data...
EXIT POLICY TOURNAMENT
Separate experiment (not included in any F/G/H total). Every real Opportunity Alert F would buy produces one common entry
(one Jupiter quote, one fill, fixed $20) cloned into S0–S4, and one common market observation per tick (a sell quote of the common tokens) that all five process.
Signal, entry, size, quote, fees, price source and monitoring frequency are identical; only the exit policy differs. Parameters frozen before the first trade. Descriptive only — no winner before 50 common executable entries.
same entry, same executable buy fill, same market observation; only exit policy differs
Waiting for data...
EXECUTION PREMIUM SHADOW
observational shadow only — not an entry rule. For every B, E and B1–B4 entry it records
execution_premium_ratio = quote_mc / signal_mc (the premium of the fresh executable Jupiter quote over the PumpDev/DexScreener reference; historical name “NO-CHASE”)
and, once the trade closes, its real executable PnL. NC10/NC15/NC20/NC30 are descriptive cohorts pre-registered on 2026-09-30: nothing is filtered, bought or sold because of them.
Only signals decided after the recorder started are counted as prospective.
Waiting for data...
Scenario details
Click a scenario to see its frozen entry and exit rules, experiment, start date and latest operations.
Waiting for data...
Scenarios
Realtime Operations
BUY shows quote → paper fill. SELL shows the trade's net result. UNEXECUTABLE shows why the order could not have been filled. Click a row for the full trade detail.
Time
Act
Scenario
Coin
Size
Signal MC
Quote MC
Market result
Exec result
Impact
Adverse
Latency
Fees
Trigger
Exit reason
Loading operations...
Scenario Comparison
Metrics only — no winner is declared. Rows are ordered by priority: executable net PnL, profit factor, expectancy, max drawdown, executability, true execution drag. Best/worst of each row highlighted.
Loading...
B original vs B1–B4 · Executable partial exits
PAPER TRADING — NO REAL ORDERS. Same entry rule, bankroll, sizing and execution model as B-CONVICTION; only the exit differs. A partial fires only when a fresh Jupiter quote for the exact partial tokens confirms the executable return. Descriptive only — no winner is declared with a small sample.
Waiting for data...
Selection Quality (F vs G vs H)
Same bankroll, sizing, execution model and exit policy: only the entry selection differs. Overlaps use the common window (F trades created since G/H started). Descriptive only — no winner is declared with a small sample.
Waiting for data...
Definition
-
Entry Rules
Exit Rules
Frozen parameters (JSON)
Execution Model
Frozen with the experiment. Calibrated on 187 real on-chain CoinLaunchr swaps (May–Jul 2026).
Remaining differences vs a real on-chain order: no real transaction is sent, so failed landings, MEV/sandwiches, RPC congestion and
order-book changes during inclusion are approximated by a fixed delay and a fixed adverse buffer; market cap for rule triggers comes from
PumpDev trades and DexScreener (which can lag a few seconds).