The Forge is a scanner. It shows you where the spread is, what it costs to capture, and what the history looks like. It does not know your position size. It does not know your jurisdiction. It does not know whether you slept last night or whether you've got the stomach for a twelve-percent drawdown on a hedged pair that temporarily unhedges. The Forge scans. The Forge does not advise. The trades are yours. The ledger is yours. The losses are yours.
// The Forge Is A Scanner, Not An Oracle
Every number on this site is best-effort. Funding rates are scraped from venue APIs on a ten-second pulse — sometimes the scrape lags, sometimes the venue flakes, sometimes the number rendered on the card is fifteen seconds stale. Every cost in the schedule is an operator estimate pulled from each venue's public fee docs at the time the file was written; your actual taker tier may differ. Every backtest is a replay of history the DB happens to have stored, which caps at the retention window and doesn't reach back to the last bear cycle. The tool is a scanner, not a crystal ball.
Nothing rendered on any page of this site is a trade order, a recommendation, or a guarantee of any kind. The Forge gives you signal. The Forge does not take your position.
// Smart Contract Risk
Every DEX on the board runs on code. Code has bugs. Bridges get exploited. Oracles get manipulated. Audits find some problems and miss others. A venue with three audits and a hundred-million-dollar bounty can still lose your position to a zero-day that nobody saw coming. This is not theoretical — Euler lost two hundred million in 2023, Multichain lost a hundred and twenty-six million, Radiant lost fifty-eight. Every DeFi protocol alive today could be a tomorrow headline.
Diversify across venues. Keep dry powder off-chain or in cold storage you actually control. Never hold more on any single protocol than you can lose overnight, because overnight is exactly when the exploits hit.
// Venue Risk
Exchanges freeze withdrawals. DEXes delist tokens mid-position. Customer service does not exist — no hotline, no arbitration, no insurance backstop. If a venue goes dark with your collateral inside, that collateral is gone. You have no legal remedy, no court to sue in, no customer agreement that survives a rug.
Venues the Forge has watched die or wobble in the last three years: FTX collapsed in 2022 and took every customer balance with it. Mango Markets lost a hundred and fifteen million to an oracle exploit that same year. dYdX forced a v3→v4 migration that stranded users who didn't bridge in time. Bluefin's public REST endpoint returned 503 for weeks on end. Drift got exploited and pulled back rate data for months while they rebuilt. Assume every venue on the board is one bad quarter from being the next headline. Act accordingly.
// Execution Risk
Slippage is a tax that scales with size. Thin books eat big orders. Multi-venue plays can fail on one leg while succeeding on the other, leaving you holding directional exposure you didn't sign up for — at the worst possible moment, in the wrong direction, with no hedge on the way. Market orders during high volatility fill at prices no scanner predicted. Gas spikes during liquidation cascades turn a fifty-dollar exit into a five-hundred-dollar exit. MEV bots front-run public mempool transactions. Every execution is a fresh roll of dice you can't see.
Leg-risk kills arb plays faster than funding ever does. You fire Venue A, the price rips three percent before Venue B fills, and now the hedge you built in a spreadsheet is a directional bet that immediately underwater. The cost table on the Forge assumes simultaneous fills. Real life does not cooperate. Slow networks, rejected signatures, gas estimator failures, wallet popups that freeze for forty seconds on a Tuesday — any of it can leave you one-legged and exposed. Size small enough that a mid-entry 5% move doesn't blow the account.
Entry and exit are the two most expensive moments of any trade. The Forge shows you the spread; it cannot guarantee the fill. Paper backtests use settled historical rates and assume perfect execution. Real execution is sloppy. Budget an extra 2–5 bps beyond the cost schedule for the gap between the scanner and the broker, and revisit the number after every trade that cost more than expected.
// Funding Flip Risk
Funding arb works until it doesn't. The spread you shorted can close, and then invert. The venue that was paying you can flip to charging you — sometimes within the same funding interval. If you're asleep when it flips and your alerts are muted, you wake up paying the tourists you were planning to collect from. Cooldowns exist for a reason. Alerts exist for a reason. Set them at thresholds you'd actually act on, and watch them when the wire taps.
Flips happen for reasons the scanner can't predict. Large liquidations on one venue skew funding toward the surviving side. New listings drain OI from established pairs. Regulatory headlines spike directional imbalance within a single candle. A venue upgrade can reset the funding calculation itself mid-trade — rare but documented. The edge that opened for a structural reason can close for a reason nobody saw coming.
Exit protocol matters more than entry protocol. Most tourists plan the entry in detail and wing the exit. Write down the exit threshold before you fire the entry: net APR below X% for Y consecutive settlements, close both legs. Net APR negative at any settlement, close both legs. Venue dark for more than an hour, close both legs. The discipline of the exit is the difference between a profitable arb and a slow bleed into the ledger.
// Collateral Risk
USDC can depeg — it did in March 2023 when Circle held reserves at Silicon Valley Bank and the peg cracked to eighty- seven cents before reserves were federally backstopped. USDT can freeze any wallet by fiat decision of the issuer. Wrapped BTC can lose its backing. Liquid staking tokens can slash. The "stable" in stablecoin is historical, not guaranteed. A hedge denominated in a broken peg is a naked position you didn't know you had until the peg cracked.
Diversify collateral across issuers. Understand the redemption mechanism of every stablecoin you hold — not just the brand. Check it before it matters, not after. Centralised stablecoins have freeze authority; decentralised ones can depeg when their backing cracks. Neither is strictly safer than the other. Both carry a failure mode; know which one each asset has, and which one you prefer to hold.
Collateral risk compounds cross-chain. A pair hedged across two venues on two chains depends on the bridge between them. Bridge exploits have pulled nine figures out of DeFi every year since 2021 — Wormhole, Nomad, Ronin, Orbit, Multichain, Poly. If the bridge securing your hedge fails, the pair that looked delta-neutral on paper becomes two isolated positions on two chains. Use bridges with real liveness track records. Verify the route before you send.
// Tax & Jurisdiction Risk
Perpetual futures trading is taxed in most jurisdictions. Some jurisdictions ban retail perp trading outright. Some permit it only for accredited or professional traders with documented net worth above a specific threshold. Some require offshore entity structures that the Forge cannot advise on. The Forge gives no tax advice, legal advice, or jurisdictional guidance of any kind.
Check your local laws before you click buy. Track every trade with timestamps, entry prices, exit prices, and funding payments. Assume the taxman will want the ledger eventually. Keep the receipts from day one.
// Tool Limitations
The Forge scrapes. Scraping has lag, cache, and failure modes. Rate limits fire. Venues change APIs without notice. Data can be stale by thirty seconds, ten minutes, or until the next outage clears. Backtests use the historical data the database has retained — which caps at a configurable retention window and does not reach back to the last deep drawdown. Portfolio pulls depend on Zerion, which doesn't cover every chain or protocol. The cost schedule in the codebase is an operator estimate; your actual taker tier, maker rebates, and slippage on thin books may all be higher than the defaults shown on the board.
Specific limits worth naming. Funding rate retention is seventy-two hours at the time of this writing — every backtest, every sparkline, every average caps at that window. Positions endpoints depend on each venue's public /info or /clearinghouse route; if a venue rotates that endpoint, the positions panel goes quiet until the fetcher is updated. Cost schedules were written to reflect public fee tiers at a point in time; venues re-tier their traders quarterly and some operators earn taker rebates that the scanner never sees. The Forge defaults to conservative numbers, but conservative is a guess, not a measurement of your account.
When in doubt, trust the venue's own UI over anything rendered here. The scanner is a first read, not a final word.
// The Mental Game
The biggest risk on every play is the operator behind the screen. Revenge trades after a loss. Size creep after a win. Averaging down into a thesis that's already dead. Ignoring the alert because the last three were false alarms. Watching the heatmap for six hours and sizing up out of boredom. Trading on tilt at 2 AM because the Forge of the Day banner caught your eye. Every professional trader has a stack of stories about the exact moment they lost discipline, and every one of those stories ends the same way.
Rules that work, written down long before the tilt hits: fixed session length (no six-hour sessions, ever); fixed max risk per trade as a percent of the ledger, enforced by the position size calculator before the order fires; no trading in the two hours after a loss above the pre-committed threshold; no trading drunk, tired, or angry. Print them. Pin them next to the monitor. The rules that aren't written down are the rules you break first.
The Forge cannot fix any of this. The warlord can. Size like you'll lose it. Hold like you'll have to explain it. Close like you mean it. The tool in front of you has no memory of last night's mistake, no fear of tomorrow's screenshot, and no stake in whether you eat well this month. That stake is yours to defend.
// The Warlord's Oath
Every warlord who lasts in the wasteland takes this oath. Rip it off. Recite it before every session. Paste it above your monitor.