LRBO3 · Povio project home

LRBO3 Backtester

A trader describes a futures strategy in plain English, the engine replays it over history, and every figure can be traced and reproduced. Start here for what is decided, what is still open, and where everything lives.

Updated 17 Nov 2026 Status pre kick-off (confirmed by sales on 16 Nov; client's written green light not found) Internal kick-off Wed 18 Nov, 12:30 PM Gašper Okršlar
Overview

About the project

About the client

LRBO3 is the holding-company name of a new Chicago brokerage; the final company name is not decided. They plan a mobile-first brokerage for futures, perpetuals, prediction markets, options and digital assets, set apart by AI features. They will be regulated, so every AI step has to be explainable to a regulator.

Why this first

Existing tools such as AskFutures ask confusing questions and return results for a different strategy than the one asked. The client wants to beat them on getting the ask right, and to prove an auditable agent framework on a small product before building more on it.

What we're building

  1. Strategy assistant: A responsive web app where a trader types a strategy or picks it from menus, answers only the questions that change the trades, and confirms a plain-English read-back.
  2. Backtest engine: A deterministic engine behind an internal API for 12 futures products and five options structures. The same strategy on the same data always gives the same result.
  3. Audit trail: A tamper-evident record per run of the request, questions, confirmed strategy, data snapshot, versions, every AI call and every figure.

The AI only translates the ask into a strategy. No number on screen comes from the model.

Pages

Open these first

Kick-off

Open questions to confirm

Nothing here is decided. Internal questions are for the 18 Nov kick-off; client questions go to the client kick-off.

Inside Povio

1
Has the client given the written green light?Sales says the project is confirmed. The email green light, the MSA and the NDA were not found. On 9 Nov there was no NDA in place.
blocking
2
Are the client's grading targets contractual?90% intent match on 50 prompts graded blind by their traders, 15 strategies matched trade for trade, 30 s median, $0.05 per backtest.
blocking
3
Do the hours hold?124 Engineer hours for the whole build, 16 of them for options pricing with implied volatility and greeks.
to confirm
4
What does design deliver, and by when?The estimate has no design hours. UI was offered as an add-on and is planned first, from the prototype, without wireframes. The client's own UX designer and the "designer joins after 15 Jan" line in the brief need squaring with this.
to confirm
5
Do we continue from the prototype or rebuild?The prototype runs on Povio hosting. The client wants it on their AWS, stood up with Terraform and Terrarium.
to confirm
6
Who is on the team, and from when?Requested: PM, architect, AI full-stack engineer, QA, UI. Engineer, QA and DevOps are not named yet.
to confirm
7
Is 24 Dec still the delivery target?Presented as four weeks of delivery plus one to two weeks to assemble the team, plus a week for design.
to confirm
8
Which feedback items are in scope?The client's 16 Nov feedback asks for a version compare view, a price chart with entries and exits, save and share, risk-based sizing and chat history. The compare view is excluded in the estimate; the others are not in the brief.
to confirm
9
When is the client kick-off, and who attends?Thursday was floated in #am-lrbo3 and not confirmed.
to confirm
10
Project adminNo Manage project exists yet. The board has no assignees, no scheduled starts and no design cards. The estimate is still In Sales and has one empty milestone, which the board inherited.
housekeeping

For the client

1
Which market data vendor and format?The brief says Chris names it at kick-off. On 11 Nov the client said they would come back on the dataset. It blocks the engine.
blocking
2
Does the vendor supply historical options chains?The five options structures need option prices, implied volatility and greeks over history.
blocking
3
Hosting requirementsChris owes the Terraform and Terrarium requirements from the 13 Nov call.
waiting on client
4
Which LLM provider, on whose account?LLM usage is not in the estimate. The cost target is $0.05 per backtest.
to confirm
5
When do the test sets arrive?The 50 prompts and the 15 reference strategies with the client's own results.
to confirm
6
Client-supplied contentUser-facing copy, disclosure text and its placement, and the exact options structure list.
to confirm
7
Engine assumptionsOrder of events when stop and target fall in the same bar, re-entries per day, and fill timing. The client wants these shown in the read-back.
to confirm
8
Strategies outside the template setThe client asked how multi-condition filters, both sides and risk-based sizing will be handled.
client asked
9
How versions and runs relate to chatsThe client asked for the data model.
client asked
10
Audit retentionHow many years records are kept, and which regulator's rules apply.
to confirm
11
Who decides on the client side?Seven people are in the shared channel. Frank sends the brief and feedback, Chris covers tech, Pete the commercial side.
to confirm
12
HandoverThe client wants to take the work in-house later: who owns the repo and cloud accounts, and what documentation they expect.
later
Hours

Budget projection

LineHoursBasis
Hours spentnot knownNo Manage project exists yet, so nothing is logged against LRBO3. Prototype and sales hours are not counted.
Hours projected: build240Estimate, task-based: Engineer 124, QA 38, PM 31, SA 31, DevOps 16. Not yet logged.
Hours projected: prompt optimization120Estimate, time and materials: 12 weeks at 8 h Engineer and 2 h QA per week. Not yet logged.
Hours projected: UI designno numberOffered as an add-on; no hours in the estimate. Left out of the total.
Projected total360Equal to the estimate baseline of 360 h. Design comes on top once it has a number.
Reference

People and links

Povio

Client

Where things live

Glossary

Terms the client uses

BacktestReplaying a rule-based trading plan over historical prices to see how it would have done.
ES / MESS&P 500 futures and the micro contract. ES is $50 per point, MES $5. NQ/MNQ, RTY/M2K, GC/MGC, CL/MCL and BTC/MBT follow the same full and micro pattern.
Continuous, back-adjusted seriesFutures expire, so history is stitched across contract months and adjusted at each roll date.
Read-backThe strategy stated in plain English for the trader to confirm before anything runs.
Silent substitutionBuilding something other than what was asked without saying so. The client's zero-tolerance item.
Strangle, vertical, iron condorStandard options structures, chosen here by delta and days to expiration.
Reference strategies15 strategies the client has implemented independently; ours must match them trade for trade.
Profit factorGross profit divided by gross loss across all trades.