All strategies

Every strategy, in one place. Results only when they are real.

The strategies Boostio runs in public, and the new ideas it is testing. Each test is written down before it runs: the rule, what it is compared with and the bar it has to clear. This is where every one of them stands today, including the ones that are blocked or have nothing to show yet.

Paper and research only: no orders, no real moneyNot investment advice4 strategies · 10 tests: 1 running, 6 collecting, 3 waiting on a missing inputFrom the tests' own records, 5-minute delay, as of 15:34 UTC
The evidence, as of 2026-11-11 15:34 UTC

Is the research learning? The honest state.

A research book weights ten research engines by their measured record and buys the top-ranked coins; a frozen twin keeps the starting weights, so the gap between the two measures the learning and nothing else. Paper money only.

The full test and its rules
  1. YesCollecting evidence

    Both books recorded on 26 days; 628 matured outcomes across 10 research engines. 5 of 10 have enough to be judged (each needs 20 over 5 data days); closest: FinRL, 8.

  2. YesChanged behaviour

    The learning rule changed the engine weights 3 times, last on 2026-11-11. The learned book now picks differently from its frozen twin.

  3. Not shownProven improvement

    Not shown. That needs the learned book ahead of its frozen twin with a 95% interval above zero at the decision read on 2027-04-16; the interim read on 2027-01-15 is a report, not a verdict.

Learned book against its frozen twin

Learned book$9,116.65
Frozen twin$9,116.65
Difference$0.00

Recorded Oct 16, 2026 to Nov 10, 2026 (26 days). Both books started with $10,000 of paper money and trade alike: fills at the live price when a decision is made, a modelled cost of 0.08% to 0.55% per side (higher for thinly traded coins), at most 5 positions of up to 20% each, each closed after 7 days or a 15% loss.

Weights v3, 3 changes. Each engine needs 20 matured outcomes over 5 data days before the rule may move its weight:

  • Qlib479 of 20
  • Wave decision policy47 of 20
  • TradingAgents36 of 20
  • FinGPT35 of 20
  • Wave trend program21 of 20
  • FinRL8 of 20
  • Wave scanner2 of 20
  • 3 other engines0 of 20

Risk and market context

Same picks, sized by risk (at most 35% each)$9,512.04
Same picks, equal dollars (the frozen twin)$9,116.65
Difference after 26 days+$395.40
Encouraging, not a verdict

Same fills, costs and exits as the frozen twin; only the split between coins differs. It counts as helping only if its largest drawdown is at least a fifth shallower than the control's and the interval of the daily return difference is not entirely below zero. First read Jan 15, 2027.

Market regimecaution
BTC, 14-day change$73,513 · -9.55%

Wave's paper account holds no BTC or ETH now: its decision policy keeps each only while it closes above its 200-day average.

Paper books only: no real money. Every figure is from the tests' own records as of the time shown. Not a forecast and not investment advice. Self-learning test · Risk-balanced sizing · All strategy tests

The strategies

What Boostio runs in public

Everything moving live: the terminal →

Wave

Paper trading

AI research and public paper trading on a simulated $100K bankroll. Every simulated trade public, 5-minute delay.

Crypto

Stock Trade Twins

Paper trading

Which rival stock leads its twin, whether the gap has measured reversion, and a shared paper strategy.

Stocks

Stock Besties

Research only

A sourced map of who funds, buys from, supplies and hosts whom in AI. Lead-lag measured forward from Oct 19, 2026.

Stocks

Paper Control Room

Paper trading

Two frozen rules on Freqtrade dry-run bots, live: every evaluation, skipped entry, paper fill and mark traced to its record.

Crypto
What each test waits for

A missing input, or time

Waiting on a missing input

Setup Boostio does not have yet. Time alone will not unblock these.

  • Kalshi copy: simple Needs trader dataA permitted source of named traders' entries and exits with their times: Kalshi's prior written authorization for leaderboard and public-profile activity, or a Kalshi-sanctioned feed with each trader's consent.
  • Kalshi copy: specialist Needs trader dataA permitted source of named traders' entries and exits with their times: Kalshi's prior written authorization for leaderboard and public-profile activity, or a Kalshi-sanctioned feed with each trader's consent.
  • Kalshi copy: consensus Needs trader dataA permitted source of named traders' entries and exits with their times: Kalshi's prior written authorization for leaderboard and public-profile activity, or a Kalshi-sanctioned feed with each trader's consent.

The Kalshi tests copy other traders on Kalshi, as separate research. Following Boostio's own trades needs none of them.

Running, waiting for evidence

Set up and recording; a verdict needs more time and trades.

  • Wave decision policy CollectingNot yet measurable: no final day since adoption. Next: first read 2027-01-15.
  • Setup-to-trend rotation RunningRunning since 2026-09-29 with $100K of paper money per book. Next: first read 2026-12-28.
  • Self-learning research book CollectingBoth books recorded on 26 days; 3 weight changes so far. Next: first read 2027-01-15.
  • Risk-balanced sizing CollectingRecorded on 26 days. Next: first read 2027-01-15.
  • Exit challengers CollectingNo result yet: the decision look needs 40 closed forward trades (an early stop for harm needs 15). Next: decision 2027-04-16.
  • AI exit warnings: Jev CollectingNo verdict before 30 closed positions with timely warnings; the record grows with every daily close.
  • AI exit warnings: Claude CollectingNo verdict before 30 closed positions with timely warnings; the record grows with every daily close.

Following Boostio's own trades

Not a test, and nothing above is needed for it. Every buy and sell of Boostio's paper books is published with its entry, exit rule and result: Wave for crypto and Stock Trade Twins for stocks. Members follow them in their desk, where a plan sizes any open house trade to their own account; any real trade is theirs to place, elsewhere. Boostio never connects to a broker or places orders, and a later copy is a different trade, at a different price.

The Wave paper recordThe Twins paper bookFollow them as a member
Strategy tests

Where every test stands

The light runs only as far as a test really is. Select a test for its page.
Test1Registered2Historical test3Forward record4First read5Decision
Wave decision policyCollectingSetup-to-trend rotationRunningSelf-learning research bookCollectingRisk-balanced sizingCollectingExit challengersCollectingAI exit warnings: JevCollectingAI exit warnings: ClaudeCollectingKalshi copy: simpleNeeds trader dataKalshi copy: specialistNeeds trader dataKalshi copy: consensusNeeds trader data
Crypto: Wave research

Rules and books that grew out of Wave's paper trading.

Wave decision policy

Collecting

BTC and ETH, each held only above its 200-day average.

Not in force.
Policy capital
$30,000
BTC half
In cash
Crypto · Paper book, in force · WAVE-POLICY-1

Setup-to-trend rotation

Running

Buy coiled leaders; sell only when the thesis breaks.

Running since 2026-09-29 with $100K of paper money per book.
Champion since start
-5.14%
BTC held
-11.79%
Crypto · Paper books · ROTATION rot-v1

Self-learning research book

Collecting

Learns which research engines to trust, against a frozen twin.

Both books recorded on 26 days; 3 weight changes so far.
Learned book
$9,116.65
Frozen twin
$9,116.65
Crypto · Shadow books · LEARN-1

Risk-balanced sizing

Collecting

Same picks as the frozen twin, sized by risk instead of dollars.

Recorded on 26 days.
Risk-balanced book
$9,512.04
Equal-dollar control
$9,116.65
Crypto · Shadow book · RISK-ALLOC-1

Exit challengers

Collecting

Two exits registered in advance, tested against the control exit.

Forward sample: 2 of 40 closed trades.
Trailing stop from the high
Rejected
Exit at the 20-day EMA
2 of 40 trades
Crypto · Exit test · TREND-EXIT-1

AI exit warnings: Jev

Collecting

Does Jev spot a weakening trend earlier than a simple rule?

18 timely Jev answers so far, 17 warnings, plus 18 backfilled answers that never count. Too early for a verdict.
Timely Jev answers
18
Backfilled answers (never counted)
18
Crypto · Shadow test · JEV-WEAKEN-1

AI exit warnings: Claude

Collecting

The same frozen question, asked of Claude every day.

27 timely answers from claude-opus-5-5 on 3 Wave paper positions so far, 22 of them warnings. Too early for a verdict.
Timely answers
27
Warnings
22
Crypto · Shadow test · CLAUDE-WEAKEN-1
Prediction markets: Kalshi copy pilot

One pilot, three ways to copy leaderboard traders.

Kalshi copy: simple

Needs trader data

Copy every move of every rostered trader.

Kalshi's market data works for Boostio's own paper testing (checked 2026-11-11): the paper fills can be priced from live quotes, fee terms and settlements. The copy books still have nothing to copy: named traders' entries and exits are shown only on Kalshi's website and app, whose terms do not allow scripted collection, and the API carries none.
Traders observed
0
Market data
Available for own use (checked 2026-11-11)
Prediction markets · Copy-trading pilot · KALSHI-COPY-1

Kalshi copy: specialist

Needs trader data

Copy traders only in the category they specialise in.

Kalshi's market data works for Boostio's own paper testing (checked 2026-11-11): the paper fills can be priced from live quotes, fee terms and settlements. The copy books still have nothing to copy: named traders' entries and exits are shown only on Kalshi's website and app, whose terms do not allow scripted collection, and the API carries none.
Traders observed
0
Market data
Available for own use (checked 2026-11-11)
Prediction markets · Copy-trading pilot · KALSHI-COPY-1

Kalshi copy: consensus

Needs trader data

Act only when independent traders agree.

Kalshi's market data works for Boostio's own paper testing (checked 2026-11-11): the paper fills can be priced from live quotes, fee terms and settlements. The copy books still have nothing to copy: named traders' entries and exits are shown only on Kalshi's website and app, whose terms do not allow scripted collection, and the API carries none.
Traders observed
0
Market data
Available for own use (checked 2026-11-11)
Prediction markets · Copy-trading pilot · KALSHI-COPY-1
How a test works

Four rules every test follows

1

Write it down first

The rule, the comparison and the bar it must clear are fixed before the first run, with a date. A changed rule is a new test, never an edit.

2

Screen it on history

Where history can say something, a historical replay can reject a rule. Passing it proves nothing: the past was known when the rule was written.

3

Record it forward

Paper books run on live prices with modelled costs, next to a control. Early numbers are a running record, not a result.

4

Read it on the date

Verdicts come at the dates set in advance, and only when an interval clears zero. Blocked or empty tests say so.

Blocked means a test cannot start until something outside it changes, such as access to data. Waiting means it is ready but has nothing new to measure. Neither ever shows an estimated or simulated result.