Skip to main content
Dequan’s referral program (→ how it works) pays referrers a percentage of their referrals’ trading fees. That kind of program is wonderful when used in good faith and toxic when farmed. So we built a defence.

​
The abuse patterns we watch for

Self-referral

Same wallet (or wallets clearly controlled by the same person) referring themselves. The system detects this through transfer-graph analysis and timing patterns.

Sybil farming

A single operator running dozens of wallets to generate fake referrals. Detected via funding patterns, behavioural signatures, IP/session correlation, and trading pattern fingerprints.

Wash farming

Two wallets trading back and forth to generate fee volume on which both earn referral rewards. Detected via trade-graph analysis.

Subscription abuse

Patterns indicating coordinated subscription gaming to harvest referral payouts beyond legitimate use.

​
How the response system works

When the system observes patterns it considers suspicious, it does not ban accounts outright. Instead it applies one of four throttle tiers:

Normal

The default state. Full referral payouts.

Soft throttle

Mild signal of unusual activity. Payouts continue but with longer settlement windows. Most flagged users return to Normal automatically once activity normalises.

Heavy throttle

Strong signal of coordinated abuse. Payouts are paused pending review. Legitimate use cases that landed here in error can be resolved through support.

Freeze

Active and clear abuse. Payouts halted, referral relationships under audit. Wallets confirmed as bad-faith are removed from the program.
The tier moves up and down based on continuing behaviour. A user soft-throttled by a noisy week of activity who then settles into normal patterns will return to Normal automatically.

​
The principle: pay real, throttle gamed

A referral program that pays uniformly to everyone makes farming inevitable, because farming is more profitable than legitimate referring at scale. A referral program that catches farming and reduces its yield makes legitimate referring the better business — which is exactly what we want. The throttle tiers are deliberate. We don’t want to ban people; we want farming to be unprofitable enough that nobody bothers. Soft throttle is usually enough — gamers move on; legitimate referrers see no impact.

​
What legitimate referrers can do to stay clean

This is almost always automatic. The system is very lenient with patterns that look real:
  • Refer real people you actually know. Friends, traders in your community, your audience.
  • Don’t auto-fund referrals from a single wallet you control. This is the classic sybil signature.
  • Don’t promise unrealistic referral earnings. Communities that overstate to drive sign-ups produce flagged sign-up patterns.
  • Let your referrals trade naturally. Coordinated wash trading between you and your referrals is detected.
99% of our active referrers never trigger any throttle, ever. The 1% who do are usually doing something they know is gaming.

​
What we will never do

  • Throttle silently without ever telling you. If you’re throttled, your account view says so and you can see why at the abstract level (e.g., “unusual referral funding patterns detected”).
  • Ban without recourse. Mistaken flags can be raised through support and corrected.
  • Use this system to throttle competitors. The system only governs referral payouts. It does not affect trading, fees, latency, or any other aspect of platform access.
  • Reveal the exact detection logic. That would let bad actors evade it. The patterns we watch for are described above; the precise weights and thresholds are not published.

​
Why this exists at all

Without a defence like this, the referral program would have to be either much smaller (defensive against farming by stinginess) or non-existent. Instead, by automating the defence, we can pay legitimate referrers meaningful, sustainable percentages that compound for them as their referrals trade. The economics are: real referrers get real money. Farmers get throttled. Everyone with good intentions wins.

Continue → How payouts work

Mechanics of the referral payout system.