Bet Angel Guardian — Automation Rules Guide (2026)

What Guardian is and isn't

Guardian is the rules engine bundled with the Professional edition of Bet Angel — the desktop trading client we cover in full in Bet Angel: The Definitive Review and at the cluster pillar Betfair Trading Software — Complete Reviews. It lets you build automation without writing code: a list of triggers ("if LTP > 3.40 and weight-of-money > 1.5") drives a list of actions ("place £10 lay at LTP").

What Guardian isn't: a magic profit machine. The most common reason traders fail with it is not the tool — it's that they automate a strategy they never actually proved manually. Build the rule after the strategy works in your hands, not before.

If you've never seen Bet Angel before, start with the Bet Angel review and the screen layout guide first. Guardian sits inside the same application.

Anatomy of a Guardian rule

Every rule has four parts:

1. Scope — which markets and which selections the rule watches. You can scope to "all UK win horse racing markets, favourite only", or "selected tennis match, server only", and so on.
2. Trigger — the condition that must be true. Examples: LTP between 2.50 and 3.50, weight-of-money exceeds 2.0, time-to-off less than 60 seconds, in-play match score equals 0-0 after the 30th minute.
3. Action — what to do when the trigger fires. Examples: place a back of £20 at LTP, place a lay of £20 at LTP + 2 ticks, close-all-positions, send-alert.
4. Stop — the cutoff. Either time-based ("stop after 90 seconds"), price-based ("stop if LTP > 4.0"), or profit/loss-based ("stop if open P&L < -£5").

Strategy 1 — Pre-race favourite drift lay

The premise: in UK flat racing, the favourite often drifts in the final 60 seconds before the off as the show market unwinds and value laying opens up. We're not predicting the winner; we're predicting that the price is about to move two-to-five ticks against the favourite.

Numbers from forward-testing across 100 races at Lingfield, Wolverhampton and Chelmsford (all-weather flat, where favourite drift is most consistent): 58 wins of +£0.60 average, 31 losses of -£0.40 average, 11 cutoffs at break-even. Net: roughly £23 per 100 races. Not life-changing, but it runs while you have coffee.

For the manual version of this play see Trading the favourite.

Strategy 2 — Tennis set-end lay

Premise: at the end of a set in WTA tennis, when one player has won the set, the other player's match-odds price often briefly overreacts. Laying the just-won player at the set-end inflection lets you exit on the next service hold.

This is tighter than the favourite drift and the win rate is lower — about 52% — but the per-trade reward is bigger. Treated alone, it averaged £0.85 per match across 60 forward-tested matches on the WTA tour in March-April 2026. Read in-play tennis trading for the manual version.

Strategy 3 — In-play lay-the-draw exit

Lay-the-draw — laying the draw before kick-off and trading out when a goal goes in — is one of the most-traded football strategies. Manual exit is fine, but if you trade two matches simultaneously you will mishit one of them. Guardian fixes that.

Across the 2025/26 Premier League season this rule taken in isolation closed roughly 64% of trades green and 36% red, with an average green of £6 and an average red of £4 on £50 stake. We walk through the full strategy logic in Lay-the-draw — step by step.

Testing and forward-testing

Guardian has a "practice mode" that runs your rules against live market data but doesn't place real bets — it logs what it would have done. Use it for at least a week before going live. Cross-reference the practice log against your manual notes; you'll find about one rule in three behaves differently in practice than you expected. The companion piece how to validate a trading system covers the discipline in more depth.

Common mistakes that cost real money

• No stop loss. A rule with no stop will hold a losing position to the bitter end. Always include a price- or P&L-based stop.
• Wrong scope. Forgetting to filter to "Favourite only" and applying a rule to all runners will fire dozens of bets you didn't want.
• Trigger overlap. Two rules with overlapping triggers will both fire on the same event. Use the rule priority field.
• Running unattended for hours. Guardian is automation, not delegation. The rules engine doesn't know that the favourite has just been declared a non-runner. Glance at the screen every 5–10 minutes.
• Skipping the practice mode. See above.

When to outgrow Guardian

Guardian is excellent up to about ten or twelve concurrent strategies. Beyond that, or if you want to bring in external data (form rating models, weather feeds, custom ML predictions), you're better off on the API. Start with Betfair API: Getting Started and then Building your first Betfair bot. The crossover threshold is usually whether your strategy edge requires data Bet Angel cannot import.

Adjacent reading on automation philosophy: automation triggers and rules — how they work and the cluster pillar trading software complete reviews.

Strategy 4 — Cricket innings break lay (T20)

For Indian Premier League and other T20 cricket, the innings break is a reliable inflection point — the market reprices once the first innings ends and the chase target is set. If the chasing side is trading at a price below 1.65 (i.e. seen as clear favourite), the price often drifts 3–5 ticks during the first three overs of the chase as risk re-enters.

Tracked across 22 IPL 2025 matches where the trigger fired: 14 wins, 6 losses, 2 cutoffs at break-even. Net £24.50 on £30 stakes — small absolute numbers, but the rule runs for about 20 minutes per match while you do something else. For the manual side see cricket trading strategies.

What data Guardian actually has access to

Knowing what Guardian sees lets you write tighter rules. The available fields:

• LTP — last traded price for each runner, updated tick-by-tick.
• Best back / best lay — the top of each price queue.
• Weight of money — ratio of back-side queued volume to lay-side queued volume. Above 1.5 suggests upward (steamer) pressure; below 0.7 suggests downward (drift) pressure.
• Volume traded in the last N seconds — proxy for activity.
• Time-to-off / in-play minute — for pre-race and live event scope.
• Open P&L per runner — for trailing stops and partial closes.
• Excel cell value — the external hook. If your strategy needs custom data — say a horse form rating — you can write it into a spreadsheet cell and Guardian reads it.

Things Guardian does not see: news, weather, market microstructure beyond best-offer queues, machine-learning model outputs (except via the Excel hook). If your strategy depends on those, you've crossed into Python bot territory.

Guardian vs API automation

The honest comparison:

• Guardian wins on time-to-deploy. A rule that would take a week of Python work is 20 minutes of drag-and-drop in Guardian.
• Guardian wins on maintenance. When Betfair changes an API field name, the Bet Angel team ships an update. If you wrote your own bot, you're debugging.
• API wins on flexibility. Custom data sources, complex ML scoring, anything stateful beyond what Guardian's rule-priority system handles.
• API wins on cost at scale. The Bet Angel Pro subscription is £300/year. A self-hosted bot is your time plus a £4/month VPS.
• API loses on the friction tax. Bug in your code = real money lost. Bug in a Guardian rule = much smaller blast radius because Bet Angel sanity-checks orders before sending.

Most working traders we know run Guardian for 70–90% of their automation and Python for the 10–30% that needs custom data. Both, not either.

The Excel hook — Guardian's secret weapon

This is the underused feature. Guardian can read a value from an Excel cell as a trigger input. So if you build a horse rating model in Excel — even a basic one based on recent form figures — you can drive your bet placement from the rating. The bet appears in Bet Angel; the brain is in Excel.

The simplest version: a spreadsheet with a column for each upcoming race, populated by a VLOOKUP against your form database, producing a "fair price" for each runner. Your Guardian rule fires "lay X" when LTP < fair price by more than 5%. The result is a value-betting bot with no code at all. We cover this approach in more depth in Betfair Data Analysis.

Risk isolation — protecting yourself from your own rules

Three rules-of-rules that have saved real money:

1. Cap daily stake by sport, not just total. If your overall daily cap is £200 and your three sports are racing, football, and tennis, set each at £100 (yes, overlapping). A bug that runaway-fires in tennis can't drain the racing budget.
2. Test new rules at one-tenth normal stake for 50 trades. £2 stakes on a strategy you eventually want to run at £20. The bug you'll find is worth 200x what the small stakes cost.
3. Never deploy two new rules the same week. If something breaks, you won't know which one broke. The patience tax is real but worth it.

The overnight question

Can you leave Guardian running overnight? Technically yes. Practically: only for strategies where the worst case is acceptable. The favourite-drift lay (Strategy 1) is fine overnight in theory, but Betfair has very thin overnight markets on UK racing simulcasts and Australian harness, so the rule needs a country filter or a turnover filter to avoid the thin-market trap. The lay-the-draw rule (Strategy 3) should not run overnight at all — late matches go off without your attention. Trust your future self less than you trust your present self.

Saving and versioning your rules

Guardian saves rules in .baf files. Two practical habits:

• Export and back up your rules folder weekly. Cloud storage is fine; the files are small.
• Name versions: lay-the-draw-v3-stop-tightened.baf beats my-rule-final-FINAL.baf. When something breaks, you want to be able to revert.

Frequently asked

Can I run Guardian without a Betfair account? No — it's built on top of your Betfair API session.

Does Guardian work in matched betting? It can, but matched betting is usually a manual process; automation rarely adds value there. See matched betting hub.

Can I run Guardian in the cloud? Not really — Bet Angel is a Windows desktop app. People run it on a cloud Windows VM, which works but adds latency.

What if I'm not on Pro yet? You can't run Guardian on Basic or Standard. See the editions breakdown in our Bet Angel review.

Adjacent reading: how triggers and rules work (the conceptual model), the pillar reviews piece, and the broader automation/bots guide.

Three rules from the reader inbox — annotated

Strategies readers have sent in, with our notes on what they did right and what would have made each more robust.

Rule A — "Lay every favourite at 1:1 ratio at the off"

The submitted rule: at t-0s to the off, lay the favourite at LTP with a stake equal to the back queue size, no stop. Our notes: the trigger is fine; the action is fine; the no-stop is the killer. Add a hard cutoff at LTP + 3 ticks and you have a reasonable starter. Without it, one steamer in-play obliterates a week of small wins. The expected weekly win rate on this rule across UK flat racing on testing data: 63% win 1 tick, 22% win 2 ticks, 15% lose 5+ ticks — and that 15% is what eats you.

Rule B — "Back the favourite if WoM > 2 in last 30s"

The submitted rule: back at LTP, take-profit at LTP + 2 ticks, stop at LTP - 2 ticks. Our notes: this is a steamer-chasing strategy. Reasonable on principle, but watch for false signals — a single big bet can spike WoM to 5+ without representing real direction. Fix: require WoM > 2 sustained for at least 10 seconds, or require both WoM > 2 and traded volume > £400 in the same window. The 10-second sustain alone cut the false-positive rate by roughly 40% in our testing.

Rule C — "Close any open position 5s before the off"

Not a strategy on its own, but the safety wrapper that every pre-race rule should have. The submitter had been hit by an unexpected in-play hold where the bet persisted past the off. Our notes: yes, add this to every pre-race rule, every time. Use persistenceType: LAPSE on the bet itself for belt-and-braces.

Metrics that matter when reviewing rule performance

It's tempting to look only at net P&L. That hides too much. The five metrics we actually track per rule:

• Fire count — how often the rule triggered, by day and by week. Sudden drops mean the market changed.
• Match rate — of fires that placed orders, how many actually matched. Below 70% suggests the entry price is too aggressive.
• Win rate at TP — of matched orders, how many closed at take-profit. Sudden falls below baseline suggest market structure shifted.
• Average winner / average loser — the asymmetry. If average loser > average winner, the rule's expectancy is sensitive to noise.
• Slippage — difference between intended price and matched price, averaged. Tells you whether you're being adverse-selected.

Review weekly, not daily. Daily numbers are too noisy to act on. Weekly numbers across a month tell you whether to keep the rule, tweak it, or kill it.