Most arbitrage software is sold on backtests and screenshots. We would rather show you live accounts. Since 20 October 2026 we have been running two strategies from the HFT Arbitrage Platform side by side on live accounts, both with a public monitoring page: the classic 2腿延迟 1 on two accounts, and the newer 3-Legs strategy on three. Same engine, same instruments, same period. The only real difference is how the trades are spread across accounts.
This article puts the two next to each other: how each one works, the live widgets (they update on their own), the numbers as of 2 November 2026, and what the trade logs reveal that the headline profit does not. The short version: both make money at the same profit factor, 2-Legs Latency 1 is more capital-efficient, and 3-Legs removes the trading pattern brokers look for first.
The two strategies in one minute
2-Legs Latency 1: the hedged two-account classic
2-Legs Latency 1 is an upgrade of lock arbitrage. Long before a signal appears, the platform opens a neutral hedge: a Buy on Account 1 and a Sell on Account 2, same symbol, same size. When the fast feed moves ahead of the slow broker by more than the entry threshold, the platform does not open a new trade. It closes the leg that is on the wrong side and lets the other leg run with a virtual trailing stop. To fix the profit and restore the hedge, it opens the opposite position again. Because one account is now flat and the other still holds the running trade, that opposite position often ends up on the same account as the running trade, so for a while one account holds a Buy and a Sell on the same symbol: a 锁.
For a deeper look at how one-account and two-account execution differ, see our guide to 单腿延迟与双腿延迟的套利.
3-Legs: a third account takes the profit-fixing leg
3-Legs starts exactly like the classic model: Buy on Account 1, Sell on Account 2, Account 3 flat. The difference appears when the profit has to be fixed. Instead of opening the opposite position on an account that already holds a trade, the platform opens it on whichever account is free, usually the third one. The Buy and the Sell of a cycle always sit on 不同的账户, so no account ever has to hold both directions on the same symbol. The full mechanism is explained step by step in our article on how rotating exposure across three accounts works.
Live monitoring: 2-Legs Latency 1
The widget below is pulled live from the independent monitoring service Strategy X-Ray, which reads the accounts with read-only (investor) access. The figures change as the strategy trades.
打开完整的 2 脚延迟 1 报告 with every account, open position and closed trade.
Live monitoring: 3-Legs
The 3-Legs group runs on three live accounts over the same period, with the same monitoring service.
Open the full 3-Legs report with every account, open position and closed trade.
Side by side: the numbers on 2 November 2026
The widgets keep moving, so here is a fixed snapshot taken on 2 November 2026, after 13 days of live trading (20 October to 2 November). Lot size was 0.10 on every trade in both groups.
| 指标 | 2腿延迟 1 | 3-Legs |
|---|---|---|
| Live accounts | 2 | 3 |
| Net profit (group) | $3,790.15 | $4,449.88 |
| Gain on the group | +19.84% | +15.42% |
| Profit per account | ~$1,895 | ~$1,483 |
| Max drawdown (group) | 0.78% ($164.30) | 1.17% ($372.36) |
| Drawdown of the accounts on their own | $1,241.60 | $2,445.11 |
| Max drawdown per account | 5.73% to 5.86% | 6.08% to 8.42% |
| 利润因子 | 1.17 | 1.17 |
| Win rate | 53.4% | 48.1% |
| Trades | 1,110 | 1,276 |
| Correlation of daily results | −0.78 (A1 vs A2) | −0.53 / −0.54 / −0.04 |
Two more measurements come from the last 200 closed trades of each group (roughly 30 October to 2 November), which we pulled from the full reports:
| Last 200 closed trades | 2腿延迟 1 | 3-Legs |
|---|---|---|
| Share of trades on XAUUSD | 97% | 98% |
| Median holding time | 13 min | 15 min |
| Trades held under 5 minutes | 24 | 14 |
| Average winning / losing trade | +$45.6 / −$50.0 | +$51.4 / −$49.7 |
| Trades split across accounts | 101 / 99 | 67 / 67 / 66 |
| Opposite positions overlapping on the same account | 64 overlaps, ~618 minutes in total | 0 |
What the numbers actually say
Same engine, same edge
Both groups show a profit factor of exactly 1.17 and trade almost only gold. That is what you would expect: the signal (fast feed vs slow broker) and the exit logic are the same. 3-Legs does not find more opportunities or better ones. It changes where the positions are held, not when the platform trades. If you are comparing the two to find "the more profitable bot", that is the wrong question.
The lock fingerprint: 64 vs 0
This is the real difference, and it is clearly visible in the logs. In its last 200 closed trades, the 2-Legs Latency 1 group had 64 cases where one account held a Buy and a Sell on the same symbol at the same time, adding up to roughly ten hours of locked exposure in about three trading days. The 3-Legs group had none over the same kind of sample. That matches the design exactly.
Why it matters: a lock on one account is one of the easiest patterns for a broker's monitoring to spot, because ordinary traders rarely hold both directions on the same symbol over and over. A two-account strategy keeps producing it by design. If you want to understand what else brokers look at, our article on 经纪人如何检测延迟套利 walks through the main signals and how brokers usually respond.
3-Legs also produced fewer very short trades: 14 trades under five minutes against 24, and a slightly longer median hold (15 vs 13 minutes). Very short holding times are another thing brokers look at, so this is a small extra benefit on top of removing locks.
Drawdown: the group hedges itself
Both groups show why these strategies are run on several accounts. Each account on its own swings by 6 to 8%, but the accounts move against each other, so the group as a whole barely draws down. 2-Legs Latency 1 drew down only $164.30 as a group, against $1,241.60 if you add up the individual accounts (about 87% less). 3-Legs drew down $372.36 against $2,445.11 (about 85% less).
The correlations explain the difference. The two 2-Legs accounts are almost mirror images (−0.78), so they cancel each other tightly. In 3-Legs, each of the outer accounts offsets the middle one (−0.53 and −0.54), but Accounts 1 and 3 are almost unrelated (−0.04). Spreading the hedge over three accounts makes the cancellation a little looser, which shows up as a higher group drawdown in both dollars and percent.
Capital efficiency: 2-Legs earns more per dollar
3-Legs made more money in total ($4,449.88 vs $3,790.15), but it also needed a third funded account. Working back from the gain figures, the 2-Legs group runs on roughly $19,000 and the 3-Legs group on roughly $29,000. Per dollar deposited, 2-Legs Latency 1 returned +19.8% against +15.4%, and about $1,895 per account against $1,483. Profit per dollar of group drawdown is also higher for 2-Legs (about 23 vs 12).
In plain terms, the third account costs you some return in exchange for a cleaner footprint. Whether that is worth it depends on how much the lock pattern costs you at your broker, and that is the question that actually decides between the two.
Which one should you run?
| Your situation | Better fit | Why |
|---|---|---|
| Limited capital, want the most return per dollar | 2腿延迟 1 | Two accounts, higher gain and lower group drawdown in this sample |
| Broker known to check for locks or hedged positions | 3-Legs | No account ever holds both directions on one symbol |
| You want accounts to last longer before they get restricted | 3-Legs | Removes the most visible pattern and gives fewer very short trades |
| You are just starting and want to learn the mechanics | 2腿延迟 1 | Simpler to set up, monitor and reason about |
| You can fund three accounts with similar spreads | 3-Legs | Similar per-account results with a much lower detection profile |
Both strategies are built into the platform, so you can start with one and switch later without changing software. For the bigger picture of the hedged, two-broker approach these strategies belong to, see our page on 对冲套利, and for the software itself, the 延迟套利软件 overview.
Read these results with care
- Thirteen days is short. The snapshot covers 20 October to 2 November 2026. A few weeks of live trading show how the strategies behave, but they do not prove long-term results.
- It is mostly gold. About 97 to 98% of recent trades were XAUUSD, which was very active in this period. Results on other instruments or in quieter markets will differ.
- The live figures will change. The widgets update continuously, so they will not match the snapshot table above by the time you read this.
- Your broker matters more than the strategy. Spreads, execution and how the broker treats arbitrage flow decide most of the outcome. The same settings at a different broker can produce very different numbers.
- No guarantee. Past results, live or not, do not guarantee future profit. Trading on margin can lose more than you expect.
常见问题
2 脚延迟 1 和 3 脚延迟 3 之间的区别是什么？
两者都使用相同的延迟信号和退出逻辑。2-Legs Latency 1运行于两个账户上，而通过定期调整利润，一个账户会持有同一标的物的买入和卖出订单（即锁仓）。3-Legs则添加了第三个账户，该账户负责确定利润调整的立场，因此账户永远不会同时持有两种方向的订单。.
在实时测试中哪种策略赚的钱更多？
3-Legs的总收益增加了（截至2026年11月2日，为 $4,449.88美元，相比之下2-Legs Latency 1的收益为 $3,790.15美元），但其使用了三个账户。按每笔存款的美元来看，2-Legs Latency 1的表现更好：+19.84% 美元，而2-Legs Latency 1的收益为+15.42%美元。两者的利润率均为1.17。.
如果策略盈利，锁仓为何重要？
Because a lock on one account is an easy pattern for a broker to spot. Once a broker flags an account, it can widen spreads, add execution delay or restrict the account, which reduces or ends the profit. Avoiding locks is about keeping accounts usable for longer.
这些实时小部件是否独立？
是的。它们来自 Strategy X-Ray，这是一个第三方监控服务，可以读取仅供投资者访问的账户信息。每个仪表板都链接到包含所有账户、开放仓位和已平仓交易的全套报告。.
我可以在一个账号上运行 2 脚延迟 1 模式吗？
Yes. 2-Legs Latency 1 can run on one account, which suits some prop-firm setups, or on two accounts, which we recommend with regular forex brokers. 3-Legs always needs three separate accounts.
我需要三个不同的经纪人来管理 3-Legs 吗？
You need three independent accounts with similar spreads on the instruments you trade. They can be at different brokers, and they should each have enough margin on their own, because the strategy does not rely on margin shared between accounts.
作者简介
