Arena Blog

A smart-money indicator paints clean zones that price seems to respect. Backtest it honestly and the magic fades. The reason is mechanical — repainting and vanished losers — not your skill.

We ran the same RSI/SMA crossover across the top 50 coins and the top 50 US stocks — identical rules, identical window. On crypto it beats the buy-and-hold benchmark 80 % of the time; on stocks, 4 %. Why a trend-following strategy works in one market and bleeds in the other.

Dual Momentum supposedly beats Buy & Hold on return AND drawdown. Our GTAA-5 backtest (2007–2026) says otherwise: not on return — but clearly on drawdown.

The 200-day line is the most-cited trend filter in markets: invested above it, out below it. We tested it as a standalone strategy across 110 assets. The result is not one-size-fits-all — on crypto it beats Buy & Hold on return AND drawdown AND Sharpe; on index ETFs it gives up almost no return while nearly halving the drawdown; on single stocks it works poorly. An honest map of where regime timing pays and where it doesn't.

Win rate alone says little. The new Details tab shows four metrics serious backtester care about: SQN, Expectancy, Profit Factor, and a monthly/yearly breakdown of every trade.

There's a way to draw Bitcoin's price that turns fifteen years of booms, crashes and mania into something almost boring: a nearly straight line.

LLMs don't trade profitably. But why would anyone expect them to? The honest answer is uncomfortable: nobody can see the future, and markets are explicitly the opposite of a forecasting machine — they're a mechanism for the ongoing negotiation of disagreement. On the Grossman-Stiglitz paradox, why one side of every trade is convinced it's smarter than the other, and why AI trading tools mathematically neutralize themselves the moment they become widespread.

Most traders writing their own backtests accidentally look into the future. The result: spectacular backtests, collapsing live performance. A look at the subtlest methodology mistake in systematic trading — from the common `shift(-N)` to the innocuous `.mean()` aggregation without rolling window — and why we manually check every Backtesting Arena strategy for bias before release.

A buy signal fires — but the unlock calendar says a large tranche of supply hits the market in nine days. Do you take the trade? Using Hyperliquid and the FDV debate as the case: what the FDV-to-market-cap ratio really means, what token unlocks empirically do to price — and the one point-in-time trap that quietly makes almost every retroactive test worthless.

Before 2021, every Bitcoin cycle had one major distribution phase in which long-term holders transferred coins to short-term holders. The current cycle broke that pattern — we've seen three separate distribution waves, each triggered by independent institutional demand shocks (ETF launch, 100k break, summer 2025 rally). An honest engagement with four possible readings, why the old 4-year heuristic is methodologically damaged, and what that means for cycle-based trading strategies.

While tech Twitter argues about USDC vs. USDT reserves, JPMorgan Kinexys quietly built a settlement infrastructure for tokenized USD deposits that handles $5B in daily volume — on Base, the same chain as x402, but 200,000× larger. A sober reading of the actually largest layer in McKinsey's three-layer monetary stack — and what the ECB's Pontes launch in September 2026 concretely means for Europe.

On Twitter, in VC pitches, in bank strategy decks, a pattern shows up: agent-payments and cross-border-remittances get sold as the same "stablecoin use case." They aren't. Different customers (software vs. humans), different volumes ($0.01 vs. $500 per tx), different compliance worlds (unregulated vs. AML-strict), different rails. Anyone wanting to serve both builds two products. Anyone who doesn't understand this loses both markets.