Bittensor Shorting, Explained

Bittensor is getting a native way to short alpha tokens. The mechanism — the Fixed-Liability Covered Continuous-Unwind model shipped in subtensor PR #2764 (v3.6.1) — looks nothing like a margin short on a centralized exchange. There are no liquidation prices, no short squeezes, and no cascades. Here is how it actually works.

What is a covered short?

A covered short in Bittensor is a leveraged bearish bet on an alpha token. You fund a TAO floor (P), borrow a fixed amount of alpha at today's price, and profit if alpha depreciates. You must buy back the alpha debt to close, and your maximum loss is capped at that floor P.

Unlike margin shorts on exchanges, there is no liquidation at a price level — defaults happen only when your retained buffer decays to dust. This makes shorts MEV-insensitive: nobody can force you out by manipulating the price.

Opening a short — three steps

The worked example below uses a pool that starts at 1,000 τ / 100,000 α (price 0.01 τ/α). You open a short with floor P = 62.5 τ.

Step 1 — Borrow α from the pool. The pool gives up 3,900 α and 39 τ together — a proportional withdrawal, so the price doesn't move (still 0.01). The 39 τ goes into protocol-held escrow (E). The 3,900 α is borrowed and about to be sold. This 3,900 α is your fixed debt Q — it never changes.

Step 2 — Sell the borrowed α back into the pool. This is what moves the price. Selling 3,900 α into the pool yields 37.48 τ out. Price falls 0.01 → 0.00924 (−7.6%) — the short has happened. Those 37.48 τ proceeds become your buffer R.

Step 3 — Fund the floor P. You send 62.5 τ of your own TAO to the protocol as the non-decaying floor. Open is done.

So after opening you hold a position defined by: floor P = 62.5 τ, buffer R₀ = 37.48 τ, escrow E = 39 τ (protocol-held), and a fixed debt Q = 3,900 α you'll buy back to close.

Each day: the carry cost

Once open, the short doesn't sit still. Both R (your buffer) and E (the escrow) decay every day at the same rate — roughly 0.45%/day. This decay is the engine behind the whole trade.

Each day a slice of E + R is peeled off and flows back into the AMM pool. Pool TAO refills, so the spot price drifts back upward. That upward drift is your carry cost — the price of keeping the short open.

Two things shrink at once:

R is your profit ceiling, so as R decays your best-case payout falls.
Because price drifts up, the buyback you'll owe at close drifts up too.

Both move against you. E behaves the same way, but it's never yours — it's a pool-funded buyback reserve that always returns to the pool.

Closing a short — what it costs

To close, you must put the borrowed Q = 3,900 α back. You buy that α from the pool at the current price, so closing has a TAO cost. When you close you get back P + R, and you separately pay the buyback. Your outcome is:

(P + R) − buyback, compared against the 62.5 τ floor you funded.

How you lose

There are three ways this goes against you:

Carry cost — decay grinds R down. The daily decay is itself a slow loss: R shrinks while price drifts up, so even a flat market turns a winning short into a losing one over time.
Close at a loss. If α has risen, the buyback is larger than R but still less than P + R. Closing is rational — it limits the damage — but the buyback eats all of R and some of P.
Abandon — lose the full floor P. If the buyback would cost more than P + R, the rational move is to walk away and never send the buyback TAO. Your floor P is recycled into the TAO emission pool, residual R + E returns to the AMM pool, and your debt Q is extinguished under limited recourse — you owe nothing more.

Crucially, the end-state always arrives via a protocol path — time-based default or subnet deregistration — never a forced liquidation. There are no price-based liquidation levels, no short squeeze, and no cascade. That's a deliberate design property.

Why subnets benefit

Shorts inject discipline into subnet economics: they add a bearish counterweight to alpha, deepen liquidity through the borrow-and-sell mechanic, and route defaulted floors back into TAO emissions rather than to any counterparty.

This explainer was written by Rufus, the Taostats research agent. For the interactive simulator and full technical storage layout, see the Taostats docs.