96 Days Later: The Wide-Band Orca Position That Never Closed Beat the Basket
In August I published I Ran the Narrow-Range Orca Strategy for 93 Days. It Lost to Doing Nothing. It covered seven simulation runs, two architecture rewrites and a lot of farmed APY. None of it beat a 50/50 basket of SOL and USDC. The cause came down to arithmetic. Every rebalance closed a position that had gone 100% single-asset at a range boundary and swapped it back to 50/50. That made the rebalancer a buy-high/sell-low ladder, and the reopen swap accounted for 99.7% of realized IL.
That post closed with a hypothesis. If the swap was the cost, the fix was to stop swapping: go wide enough that you rarely exit, never force-close, and let fees accrue.
On August 10 I deployed exactly that, again in paper only, against live SOL pricing and live Orca SOL/USDC 0.04% pool state. This is the 96-day check-in. SOL is now trading below our entry.
Capital at risk: still $0. Every figure below is simulated. I'll keep repeating that until it stops being true.
The Setup
- One position. The old setup ran a stack of bands at different widths; this one runs a single position.
- Band: $40.34 → $145.62, centered geometrically on about $76.6 SOL at entry. That's the ±90% configuration, the widest tier in the original width sweep. In that sweep it was one of only two widths, along with ±60%, that beat a 50/50 basket in all four regimes tested.
- No forced closes. There's no out-of-range trigger, no grace timer and no ATR regime switching. If price leaves the band, the position just sits there 100% in one asset until price comes back.
- No reopen swap, which removes the leak the previous post found.
- Benchmark: a $1,000 50/50 SOL/USDC basket marked at the same timestamp. APR isn't the benchmark, and neither is in-range %. The previous post explains why both of those metrics flatter you exactly when you're losing.
The Scorecard
| Metric | Value |
|---|---|
| Days deployed | 96 |
| Starting capital | $1,000.00 |
| Position mark-to-market | $1,036.67 |
| 50/50 HODL basket | $994.76 |
| Excess vs basket | +$41.91 (+4.21%) |
| Fees earned | +$41.96 |
| Impermanent loss (open, unrealized) | −$0.05 |
| Price P&L (shared with the basket) | −$5.24 |
| Rebalances | 0 |
| Gas + slippage | $0.00 |
| Fee / IL coverage | n/a (no material IL) |
| Ledger reconciliation | equity = mark, $0.00 residual |
The P&L breaks down into three lines with no fudge factor: $1,000 + $41.96 fees − $0.05 IL − $5.24 price = $1,036.67. The reconciliation check runs on every tick. Two months of the original dashboard showed a ratcheting principal figure that could only go up, and I'm not shipping another number that hasn't been reconciled against the ledger.
On a simple run-rate basis, +4.21% over 96 days annualizes to roughly 16% excess over the basket. Treat that as indicative only. Ninety-six days in one regime isn't a yield you can underwrite.
"SOL Is Down." What That Did to the Position
SOL marks at about $75.80 at the time of writing, roughly 1% below the band's center at entry. The position sits about 34% of the way up the band, still well inside it.
Being down 1% from entry also means price has round-tripped. Price spent much of the period away from where we opened, and the position's IL tracked that. Thirty days ago the position was about $20 behind where it is today relative to the basket. That shows in the trailing windows:
| Window | Excess vs basket |
|---|---|
| Last 7 days | +$16.49 |
| Last 30 days | +$62.12 |
| Since inception (96 days) | +$41.91 |
The 30-day excess is larger than the lifetime excess. That isn't an error. Over the last month the position earned fees, and it also recovered the unrealized IL it had built up while SOL traded away from entry. That's the "impermanent" in impermanent loss doing exactly what the name says, which can only happen if you don't crystallize it. Under the old rules, every one of those excursions would have triggered a close and a 50/50 reopen swap, and the IL would have been realized for good.
The 7-day print is also telling. $16.49 in one week against a lifetime average of about $0.44/day: volatility drives volume, and volume drives fee income. On a wide band you collect that flow without getting rotated out of the position.
Why It Worked: No Swap, No Ladder
Here's what the wide band did differently from the 93-day experiment:
- It never paid the per-close toll. The old data showed IL per close was flat at about $9–10 regardless of hold time, while fees grew with time in range. With zero closes, the position paid zero tolls.
- It never sold SOL at a local low or bought it at a local high. The swap ladder simply doesn't exist here.
- Unrealized IL was allowed to mean-revert. A narrow band gets knocked out by the excursion; a wide band rides it out and earns fees the whole time.
- Friction was effectively zero. One open, no swaps, no gas drag. A narrow-range strategy has to clear that friction on every cycle.
The trade-off is the one the band-width table predicted. Per dollar of TVL, a ±90% band earns far less than a ±1.25% one. But fee income across widths varied only about 3× in the backtest, while IL varied about 4× and was two to seven times larger in absolute terms. Giving up some fee density to cut IL exposure that much is the right trade.
What This Does Not Prove
This is a positive result, so it gets the same scrutiny as the negative one.
- This was the friendliest possible outcome for an LP. Price finished near entry. Concentrated liquidity is short gamma, and a short-gamma position always looks good when the underlying round-trips. The real test is a sustained directional trend, especially a breakdown below $40.34, where the position goes 100% SOL and rides it all the way down. That scenario hasn't happened in this run. In the 233-day backtest that included the 2025 crash, ±90% finished +$58.77 over the basket but still down in absolute dollars ($871.37 on $1,000). Beating the basket isn't the same as making money.
- 96 days is one sample. It shows the design behaves as intended in one regime. It doesn't establish an edge.
- Fees are modeled from live pool volume times our modeled liquidity share. They aren't on-chain receipts. A $1,000 position is small enough that it shouldn't move the pool, but slippage and fills on a real reopen or claim haven't been tested.
- Compounding in the sim is simulated too. One $25 compound event was booked. That's fee income reinvested, not profit, a distinction the last post spent a whole section on.
- Out-of-band behavior is the open risk. "Never force-close" means that below the band you're simply long SOL with no fee income. That's a deliberate choice, and it's the cost of not running the ladder.
Where This Leaves the Playbook
If you LP on Orca SOL/USDC, here's what the combined 189 days of paper data support:
- Width matters more than any tuning parameter. Get the band right first. Regime, then width, then whether you close at all, then reopen policy. Grace windows, slippage tweaks and compounding cadence barely move the result.
- Go wide. In our data, ±60% to ±90% was the only zone that held up across regimes.
- Don't auto-rebalance back to 50/50. If you must reset a position, re-enter one-sided with the asset you're already holding instead of market-swapping at the boundary.
- Score against the basket, not the APR. Track excess vs a 50/50 HODL and the fee/IL coverage ratio. Fee APR and in-range % are diagnostics, and both peak at the worst possible times.
- Expect a slow, boring equity curve. A position that books about 4% excess in a quarter with zero trades is what working looks like here. A position printing 255% APY is usually funding someone else's exit liquidity.
What's Next
The 96-day Phase 1 window is complete and the paper gate has passed: excess over the basket is positive, the ledger reconciles, and no realized IL leaked through. The next step is a small-capital mainnet deployment of the same ±90% configuration, so we can replace modeled fees with on-chain claims and real execution costs. I'll publish those numbers here, including if they come in worse.
One more benchmark question this run raises: nobody holding SOL long-term has to hold it idle. A fairer passive comparison is probably a basket built on a liquid staking token (JitoSOL, mSOL) plus USDC, which earns validator yield for doing nothing. That's the benchmark we plan to score this position against next, and if part of the edge turns out to be unclaimed staking yield, I'll report that too.
The full background, including all seven failed runs, the swap-ladder proof and the band-width sweep, is in the original post: I Ran the Narrow-Range Orca Strategy for 93 Days. It Lost to Doing Nothing.
Nothing here is financial advice. It's a public lab notebook. Concentrated liquidity is a short-volatility position with real downside, and paper results don't include the execution risk you'll take on in production. DYOR, size accordingly, and don't deploy capital you can't afford to see marked down in a trend.