We currently run 45 paper trading bots with a total allocation of 45,000 dollars. Each bot receives 1,000 dollars. This is deliberately simple, and it is deliberate that it is simple. Equal allocation during paper trading lets us isolate strategy performance without introducing allocation bias. When we transition to live trading, the capital allocator switches to risk parity weighting. But the paper trading phase is about collecting clean performance data, and equal allocation is the cleanest way to do that.
The Full Bot Breakdown
Our 45 bots span six distinct strategy types, three timeframes, and four data source categories. Here is exactly what is running.
Mean Reversion Bollinger Bands (15-minute) -- 13 bots. Six bots trade the original validated symbols (SHIB, DOGE, AVAX, SOL, LINK, SUI) with parameters bb_period=30 and bb_std=2.5. Seven bots trade the newer symbol expansion (NEAR, ARB, PEPE, WIF, OP, APT, INJ) with bb_period=48 and bb_std=2.5. The parameter difference comes from Phase 2 optimization, which found that longer Bollinger Band periods work better on the newer symbols. Together, these 13 bots represent our highest-conviction strategy on our highest-conviction asset class: high-beta altcoins that oscillate around a moving average.
Momentum RSI+MACD (15-minute) -- 5 bots. Trading SUI, SOL, AVAX, LINK, and DOGE with rsi_period=10 and sma_period=10. These are the symbols where momentum validated with Sharpe ratios between 3.5 and 7.8 across five regime periods. Momentum captures trending moves on the same altcoin basket where mean reversion captures reversions. The two strategies are naturally anti-correlated on the same symbols, which provides genuine portfolio diversification.
Momentum RSI+MACD 4-hour -- 6 bots. Trading ETH, BTC, SOL, ADA, SHIB, and AVAX with rsi_period=10, sma_period=10, rsi_oversold=35, and rsi_overbought=65. This is our first viable strategy for BTC and ETH. Fifteen-minute strategies fail on these assets because they are too efficient at short timeframes. The 4-hour timeframe with relaxed RSI thresholds (35/65 instead of the standard 30/70) captures the larger swings that BTC and ETH do exhibit.
Leverage Composite (1-hour) -- 3 bots. Trading ARB, OP, and WIF with oi_lookback=14, oi_threshold=0.024 (0.03 for OP), funding_extreme=0.00036 (0.0003 for OP), and lsr_crowded_long=0.65. This is our first derivatives-based strategy, using open interest changes, funding rate extremes, and long/short ratio crowding as signals. It was validated across three sub-periods (October-December 2025, January-February 2026, March-April 2026) and was profitable in all three.
Correlation Regime (4-hour) -- 6 bots. Trading NEAR, DOGE, AVAX, MATIC, ADA, and DOT with corr_threshold=0.5, price_sma_fast=5, price_sma_slow=20, and btc_price_sma=14. This macro strategy detects shifts in the BTC-altcoin correlation structure and trades accordingly. It validated as ROBUST on all six symbols with 4 out of 5 periods positive for each.
NUPL Cycle Filter (4-hour) -- 7 bots. Trading INJ, LINK, TRX, UNI, AVAX, BTC, and DOT with euphoria_threshold=0.75, capitulation_threshold=0.0, trend_lookback=7, and use_sopr=false. This on-chain strategy uses Bitcoin's Net Unrealized Profit/Loss to identify market cycle zones and trade altcoins accordingly. It was the strongest on-chain strategy in validation, with INJ and LINK achieving 5 out of 5 periods positive.
Stablecoin Supply Momentum (4-hour) -- 5 bots. Trading AVAX, NEAR, SOL, LTC, and ATOM with ema_period=14, roc_lookback=14, and roc_threshold=0.005. This on-chain strategy tracks the rate of change in total stablecoin market cap as a proxy for capital inflows to crypto. When stablecoin supply accelerates, it signals new money entering the market.
Why Equal Allocation During Paper Trading
The temptation is to allocate more capital to strategies with higher backtest Sharpe ratios. Mean reversion Bollinger Bands on PEPE achieved a Sharpe of 19.25 in backtests. NUPL cycle filter on INJ achieved a Sharpe of around 1.45. Why give them the same dollar allocation?
Because backtest Sharpe is not live Sharpe. Every strategy we deploy has been validated across multiple regime periods, but validation tells us the strategy has an edge, not how large that edge is in current market conditions. Paper trading is the phase where we measure live edge. Biasing capital toward high-backtest-Sharpe strategies before we have live data would bake the overfitting risk directly into the portfolio.
Equal allocation also simplifies performance comparison. When every bot has 1,000 dollars, comparing dollar PnL across bots is comparing strategy performance. No need to normalize for allocation size or calculate risk-adjusted returns. Bot 7 made 43 dollars and Bot 22 lost 18 dollars. The comparison is immediate and honest.
The Transition to Risk Parity
When we move to live trading, the capital allocator will switch from equal weighting to inverse-volatility weighting (risk parity). The principle is straightforward: each bot should contribute equally to portfolio risk. A high-volatility bot needs less capital to produce the same risk contribution as a low-volatility bot.
The formula calculates the inverse of each bot's recent return volatility, normalizes these to sum to 1.0, and multiplies by total capital. A bot whose returns have a standard deviation of 2 percent per day gets half the allocation of a bot whose returns have a standard deviation of 1 percent per day.
Two constraints prevent extreme allocations. The single-bot cap at 40 percent ensures no bot dominates the portfolio even if it has unusually low volatility. Surplus capital from capped bots redistributes to uncapped bots using the same inverse-volatility weights, so every dollar stays deployed.
Capital Efficiency Across Strategy Types
The six strategy types have different capital efficiency profiles that will matter for live allocation.
Mean reversion bots on 15-minute timeframes trade frequently (multiple signals per day) with small position sizes and tight stops. Each dollar of capital sees high utilization. Momentum bots on 15-minute timeframes trade less frequently but hold positions longer, so capital is tied up in open positions more of the time.
The 4-hour strategies trade the least frequently. A 4-hour momentum bot on BTC might execute 2 to 5 trades per week. Capital sits idle most of the time. This low utilization means 4-hour bots need proportionally more capital to generate meaningful absolute returns, but they also contribute less risk per dollar because they are less active.
Derivatives and macro strategies occupy a middle ground. The leverage_composite bots on 1-hour timeframes trade several times per week. The correlation_regime and NUPL bots on 4-hour timeframes trade similarly to other 4-hour strategies but their signals are based on fundamentally different data sources, which provides the diversification benefit that justifies their capital allocation.
The Numbers Going Forward
At 45,000 dollars total with equal allocation, each bot's 1,000 dollar position is small enough that slippage is negligible even on mid-cap altcoins. The total portfolio exposure cap of 50 percent means a maximum of 22,500 dollars can be in open positions at any time across all 45 bots. In practice, not all bots are positioned simultaneously, so actual exposure typically ranges from 15 to 35 percent of total capital.
For live trading, the total capital will increase but the allocation logic stays the same. The critical decision is not the total dollar amount. It is the allocation methodology: equal for paper trading to collect clean data, risk parity for live trading to equalize risk contribution, with hard caps to prevent concentration regardless of what the math suggests.