Two stocks. One spread. Reduced market exposure.

Trade the relationship. Not the forecast. CounterPoint builds systematic trading systems that take the market’s direction out of the trade. The flagship is an equity pairs system designed to reduce broad-market exposure: long one stock, short a related one, and trade only the distance between them. It is traded live by its developer, in his own account, with real money — the same rules, the same code, no separate “demo” version. It is designed to reduce broad-market exposure, not eliminate risk. “Market-neutral” describes the intended long/short construction, not a guarantee of zero market exposure.

  • Live at IBKR with real money — the author trades the flagship in his own account, and runs it live every trading day.
  • Institutional pedigree: strategies passed CloudQuant’s institutional review, were allocated $17M of firm capital to trading strategies for company profit, then licensed.
  • Published research on pairs-trading holding periods.
THE SPREAD · Stock A vs. Stock B · z-score

The pair on one line: enter at ±2.5σ, exit at ±1.0σ. Illustrative diagram — not a real pair or a track record. The full chart is below.

ENTRY BAND ±2.5σ / ±3.25σEXIT reversion to ±1.0σIllustrative chart
Strategy
Equity pairs
Exposure
Designed to reduce broad-market exposure
Execution
Fully systematic
Status
Live — developer’s own account
THE SPREAD CHART · Stock A vs. Stock B · z-score of the spread
Move your cursor across the chart to read the spread
Spread (z-score) Entry / exit markers Entry band ±2.5σ · Add band ±3.25σ · Exit band ±1.0σ ENTRY BAND ±2.5σ / ±3.25σ · EXIT reversion to ±1.0σ · Illustrative chart

Illustrative diagram using generic Stock A and Stock B. It shows how the system sees a pair — it is not a real pair, a track record, or a performance record.

The Flagship

One spread. Four decisions.

Two related stocks tend to move together — but not in lockstep. Day to day, one runs ahead of the other, then the gap closes. The flagship does not predict which stock wins or where the market goes. It measures the gap between a pair, waits until that gap is stretched far beyond its normal range, and takes the other side of the stretch: long the laggard, short the leader, in matched size.

STEP 01

Pair stocks that belong together

Start with stocks whose prices have a durable, measurable relationship — the kind that wander apart and come back, rather than drifting away for good.

STEP 02

Chart the spread, not the stocks

The trade lives on one line: the spread between the two, standardized as a z-score so “stretched” means the same thing on every pair.

z = (spread − mean) / σ
STEP 03

Enter only at extremes

No stretch, no trade. A first position opens only when the spread reaches the outer entry band — a dislocation, not a hunch.

First entry at ±2.5σ
STEP 04

Exit on the way home

The position closes as the spread normalizes — it does not need to cross zero or overshoot the other way to have done its job.

Exit at ±1.0σ
The Method

Designed to reduce broad-market exposure

Every position is two positions: a long and a short, sized against each other. “Market-neutral” describes this intended construction, not a guarantee of zero market exposure — residual beta, sector, factor, borrow, liquidity, gap, and execution risk remain. If the whole market falls, the short leg is built to offset the long leg. If the whole market rallies, the same offset applies in reverse. What is left — the part the system actually trades — is the relative move between the two stocks.

In plain terms: the spread can stretch in either direction, and the system takes the opposite side of the stretch — long the laggard and short the leader, or the reverse, depending on which side stretched.

That construction is the point of the name. In music, counterpoint is two independent lines written against each other so the interest lives in how they move relative to one another. A pairs book works the same way: neither leg is the trade. The relationship is the trade.

The rules are fixed in advance and executed by code — pair monitoring, signal checks, entries, and exits all run on schedule through the trading day. Discretion does not get a vote at 2:47 in the afternoon.

MARKET LONG ONLY LONG + SHORT PAIR
When the market falls, a long-only position falls with it. A matched long/short pair is built so the market’s move mostly cancels — what remains is the pair’s own spread.

The flagship rule set, in plain terms

First entryThe spread reaches the entry band
±2.5σ
Add levelIf the dislocation deepens, size steps up once
±3.25σ
ExitThe spread normalizes back toward its mean
±1.0σ
Hard stopA pre-set limit per pair ends the trade if the relationship breaks instead of healing
FIXED IN ADVANCE
Book limitsA capped number of pairs at a time, spread across unrelated names
DIVERSIFIED
Method Specification

How the flagship is calculated.

No source code is published here. The rules below are precise enough to understand what is being licensed. Position sizing is equal dollars per leg, capped and floored. The spread is standardized using the 60-day standard deviation of spread log returns. Entries are on daily closes at 2.5σ (first level) and 3.25σ (add level), the exit is at 1.0σ, and the maximum hold is 50 days. A pair must have correlation of at least 0.50 and cointegration Spot of at least 0.90 (tier A) or 0.88 (tolerance tier), using the longest window of 730 days, otherwise 504 days. Half-life is a flag only, not an entry gate. Pairs are selected by cointegration and correlation.

Who Builds This

Serious about what a claim means.

David has been a professional algorithmic systematic trader and system developer for over 25 years, since 2000. Strategies he has developed passed CloudQuant institutional review, were allocated $17M of firm capital to trading strategies for company profit. CounterPoint is the work of one developer-trader: an ex CloudQuant trader whose strategies passed institutional review and were licensed — and who now runs his own systems, in his own account, where the results are his own to live with. The flagship on this page is not a backtest looking for a buyer. It is a working system, running live, carrying real risk.

Over 25 years, since 2000

Professional algorithmic systematic trader and system developer for over 25 years.

$17M of firm capital

Strategies passed CloudQuant institutional review, were allocated $17M of firm capital to trading strategies for company profit, and traded on firm capital for company profit.

Live, with his own money

The flagship pairs system runs in the developer’s own brokerage account. No demo mode, no paper book standing in for the real one.

Why This Exists

Rules you can inspect. Risks we name.

CounterPoint is built for the trader who wants systematic execution without buying a black box or writing a codebase. The flagship is a focused tool, not a managed account and not a promise of returns.

Built by a systematic trader

David has developed algorithmic systems since 2000. Prior strategies passed CloudQuant institutional review and were allocated approximately $17M in trading capital. This is self-reported context — verify the details before relying on it.

Run by its builder

The flagship is traded by its developer in his own IBKR account. That establishes live use; it does not by itself prove future performance.

● LIVE — NOT A TRACK RECORD

No other systems offered

Only the flagship is offered. Nothing else is listed here, and no other system is included in the flagship lease.

Systems

What exists, and what doesn’t — stated plainly

One flagship, traded live. Everything else is listed with its real status. Nothing on this page is for sale today, and there are no purchase links because the product is not ready to be sold.

SystemWhat it isStatus

CounterPoint Pairs — Flagship

Equity pairs designed to reduce broad-market exposure. Long/short, spread-based entries at ±2.5σ with a ±3.25σ add level and a ±1.0σ exit, executed systematically.

● LIVE — TRADED BY THE DEVELOPER
Not yet offered for licensing
Pricing

One flagship. One clear lease.

The flagship is not yet offered for licensing. Licensing opens only after paper testing is complete. The terms below are the planned launch pricing. There are no payment or purchase links on this page.

Flagship · Annual lease · Not yet available

CounterPoint Pairs — Flagship

Market-neutral equity pairs system. Annual lease, per licensed account.

$525 / year

Indicative launch price — not currently orderable.

Standard annual lease, all buyers$525 / year
First-buyer discountNone — $525 / year for all buyers
Phone notifications includedIncluded

Phone notifications included in this flagship lease: entry, exit, entry failed, and exit failed alerts, delivered via ntfy to the buyer’s own private topic. Alerts run only while the lease runs — they lapse when the lease lapses. Join the flagship waitlist below to be notified when licensing opens.

Join waitlist

Waitlist only — not for sale yet.

Waitlist only at this stage. No payment is taken and no license is issued until paper testing is complete and licensing is announced here.

Capital, guardrail, and hard stop

At $10,000: gross guardrail $19,200, leg cap $3,240, hard stop $240 per position record. At $25,000: gross guardrail $48,000, leg cap $8,100, hard stop $600 per position record.

What you receive

Installer, license key, rule summary, Interactive Brokers connection, and phone alerts included (entry, exit, entry failed, and exit failed). Support is 1–2 business days at david@counterpointtradingsystems.com. What is not included: a brokerage account, market data, a Windows machine or VPS, and trading advice. Update, renewal, and cancellation terms are not stated here and are not invented in this summary.

Performance, plainly: No independently verified performance record is being offered for the retail system on this page. The live-trading statement is a description of the developer’s own activity, not a performance claim.

Also planned — not yet available

The following is planned only. It is not yet available.

Second-machine license

An additional license for running the flagship on a second machine, for the same buyer, with the same expiry, only after a full $525 purchase.

PLANNED · $195 / YEAR
Performance — Hypothetical

Read the boundary before the promise.

Hypothetical backtest, not a live record, not retail trading results. These are raw performance numbers without filtering: all completed layers are included, no cherry-picking, wins and losses both counted, drawdowns shown. No independently verified performance record is being offered for the retail system on this page. The live-trading statement is a description of the developer’s own activity, not a performance claim.

Results — $25,000 start, $1 per leg per side

RunLayersWinProfit factorCompleted returnMax drawdown
YTD 2026, fixed size
Jan 2 – Oct 29, 2026
5377.4%2.43+1,894 bps (18.94%)-3.50%
YTD 2026, compounding5477.8%2.43+2,238 bps (22.38%)-4.07%
5 years, fixed size
Oct 28, 2021 – Oct 29, 2026
30274.8%2.52+10,665 bps (106.65%)-4.39%
5 years, compounding30876.3%2.62+21,280 bps (212.80%)-6.99%

Basis points, in plain terms: 100 bps = 1%. So +1,894 bps is +18.94%.

Rules modelled

174 pairs. Entries 2.5σ and 3.25σ, exit 1.0σ, 50-day maximum hold, correlation at least 0.50, cointegration Spot at least 0.90 tier A / 0.88 tolerance, $10 floor on either leg, 12% broken-pair entry block and exit, 6 slots maximum, $1 per leg per side, fills at close, exits before entries.

Limits — these stay with the numbers

  • Earnings blackout, corp-event, and M&A exits not modelled — profit is overstated to that extent.
  • Live manual blacklist not modelled.
  • No VIX freeze at 30.0 modelled.
  • No borrow financing, no slippage beyond $1 per leg per side.
  • Survivorship: the current 2026 universe run back to 2021; dead or dropped tickers are absent.
  • Adjusted closes, daily close passes only — live runs every 5 minutes in market hours.
  • Stricter than live fuzzy in one way: this run requires a Spot pass; live fuzzy can trade a Spot-fail pair if fuzzy is GO, so this is not an exact live replication.
Waitlist open

Join the waitlist — get notified when it is ready

Waitlist only — not for sale yet.

Counterpoint is launched as a waitlist first. No product is for sale yet, no payment is taken, and no license is issued until paper testing is complete. Joining the waitlist means you will be notified first when licensing opens, like a restaurant texting you when your table is ready.

Manual paper trial, when it opens: submit your details, the request is reviewed, and an installer and trial key are emailed back. There is no automatic download and no live-account trial. Response is 1–2 business days. At trial expiry, entries stop and exits continue. The software uninstalls cleanly, and logs stay local on your machine.





Any format works, for example 678-656-3853, (678) 656-3853, or 6786563853. US numbers only for texts.


Joining the waitlist is not a purchase, not a trial, and not a license. You will be contacted only about Counterpoint licensing, and you can ask to be removed at any time.

Privacy, plainly: your email and phone number are never sold, never rented, and never shared with anyone else. They remain private, used only to notify you when Counterpoint licensing opens. No spam, no marketing lists, no third-party messages. Reply STOP to a launch text, or email to be removed, at any time.

System Requirements

What the system does. What stays with you.

CounterPoint Pairs is a Windows-only product you run on your own machine, connected to your own Interactive Brokers account. It does not run on every computer, and the installer cannot fix an unsupported machine for you. If you do not meet this spec, do not trial it on that machine.

CounterPoint automates

  • Pair monitoring and scheduled signal checks
  • Entry, add, and exit rules
  • Order submission through your IBKR connection
  • Position checks and rule-based book limits
  • Status signals for clean, stale, error, or halt states

You remain responsible for

  • Funding, permissions, market data, and borrow access
  • Keeping Windows, TWS/Gateway, and the connection available
  • Monitoring the account during market hours
  • Manual intervention if a leg, connection, or order fails
  • Position size, risk capital, and whether to continue
  • Windows 11 recommended, 64-bit only. Windows 10 technically works but is out of support as of October 14, 2025, use it at your own security risk. No Windows 7, no Windows Vista, no Linux, no Mac.
  • Interactive Brokers TWS or IB Gateway, installed and logged in. Paper account for trial, live account for live. API access enabled. For live, Read-Only API must be unchecked.
  • Ports, or your actual configured port: paper TWS 7497, paper Gateway 4002, live TWS 7496, live Gateway 4001. Your firewall must allow that port.
  • Market data for the stocks traded, and a margin account that allows short selling. Confirm data and shorting with Interactive Brokers first. A retirement account that cannot short is not suitable.
  • Execution is automatic. In live trading you cannot inspect every order before it is sent; the system places orders itself on a pass.
  • Admin rights on your own machine, to install and to create the scheduled task.
  • A machine that stays on and stays connected during market hours, weekdays 09:30–15:55 Eastern. If the machine or TWS is off, the software cannot trade or exit.
  • Minimum $10,000 in the account, $25,000 default sizing.

VPS, desktop, or laptop

A Windows VPS is fine, and is often the better choice. A VPS stays on, does not sleep, and does not close its lid. It still has to meet every item above, with TWS or Gateway running on that VPS, and the license bound to that VPS machine.

A desktop is fine if it stays on during market hours with TWS logged in. A laptop is the weakest choice for a scheduled system. A laptop that sleeps, closes, loses Wi-Fi, or runs out of battery will miss passes, including exit passes. If you use a laptop, it must stay awake, plugged in, and connected during market hours, or use a VPS instead.

Launcher, status, and end-of-day Stop — planned, not yet in the current installer

Do not expect this to run invisibly 24 hours a day, 5 days a week, on the same machine you use for everything else. That machine sleeps, reboots, loses Wi-Fi, and gets TWS closed, and every one of those breaks a pass, including an exit pass. That is why a Windows VPS is often the better choice.

Launcher: one double-click in the morning, "Start Counterpoint Pairs." It checks TWS and your license, then starts the scheduled task and a clean green, red, or gray status indicator, not a black terminal window. Green means a recent clean pass in market hours, red means stale, an error, or HALT, gray means outside market hours. End-of-day Stop, if you wish: one double-click after market close, 15:55 Eastern. It disables the scheduled task and stops the indicator, so no new pass starts. It does not close open positions, they stay open at the broker overnight and resume next morning on Launcher. Stop is optional, if your machine stays on, leave it running.

Trade timing: passes run at 09:30, 09:35, 09:40, and so on, every 5 minutes to 15:55 Eastern. A trade does not fire at exactly 09:30:00, it fires during a pass, typically a minute or two after that pass starts, if a pair qualifies. If a pass fails or no pair qualifies, the next chance is the next 5-minute pass.

Licensing limits, machine by machine

  • One license, one buyer, one machine, one year, bound by a fingerprint from the Windows machine identifier and disk identifier. Paper and live use the same key for the same buyer. A second license at $195 is a second machine, not a second account, only after a full $525 purchase, with the same expiry.
  • Moving machines is a machine change, even for the same buyer: laptop to VPS, VPS to laptop, or VPS to VPS. That uses your one free transfer, or requires a second license at $195 a year for concurrent use on a second machine, only after a full $525 purchase, verified by order number, with the same expiry as the first.
  • One license does not cover two machines at once. Running the same license on a laptop and a VPS at the same time breaches the license. Transfer is contractual, not technical: there is no server to disable the old machine, so you must stop using and remove the old installation when you transfer. Consequence: if concurrent use is found, renewal is not offered. The one free transfer is enforced by the vendor: a new machine needs a newly signed key from the vendor, recorded against your order number, and a second transfer without a $195 second license is not issued.
  • A VPS rebuild, snapshot restore, disk replacement, or Windows reinstall that changes the machine identifier can break activation, even with the same VPS provider. Tell support before you rebuild a VPS, and expect a transfer or re-activation to be needed.

Trial request checklist — check every box to request a trial

This is a manual trial request, not an automatic download. Check every box, enter your name, email, and paper account, and the request button will open a pre-filled email to david@counterpointtradingsystems.com. Send that email, and a trial key and installer will be sent back to that email address by hand. A fake email never gets a key, and without a key the software will not run. There is no public download link on this page. Trial is 14 days, paper account only, then buy. There is no live trial. This is a manual paper trial: submit your details, the request is reviewed, and an installer and trial key are emailed back. There is no automatic download and no live-account trial. Response is 1–2 business days. At trial expiry, entries stop and exits continue. The software uninstalls cleanly, and logs stay local on your machine.

Check every box above and enter a valid email address to enable the request button. The button opens a pre-filled email to the vendor, it does not download software directly.

Reliability

Trading software needs failure language.

Existing today: On a partial fill, the system cancels the pending leg, reverses any filled legs, verifies the reversal, and records the failure; 2 underfills in 24 hours blocks that pair. HALT blocks new entries but exits still run. On restart or reconnect, the system recovers and reconciles positions against the broker. Stale data and rejected orders are logged. Stop pauses new passes. If the machine or connection fails, you must be able to close a position by hand in TWS.

Planned, not existing: a post-pass quantity check, a pre-placement size check, and a 30-second rejected-order wait. These are not claimed as current features.

Research Notes

Notes from the desk

Plain-English explainers on how systematic pairs trading actually works — the charts, the signals, and the reasoning — written for traders who want to understand the machinery, not admire it.

Reading a Spread Chart: How Pairs Traders See a Trade

A stock chart shows you a price. A pairs trader barely looks at it. The chart that matters shows the distance between two stocks — and learning to read it is learning to see the trade itself.

Start with two stocks that move together

Take two stocks in the same line of business — call them Stock A and Stock B. Same industry, same customers, same weather. Most days they rise and fall together, because the forces pushing on one are pushing on the other. Neither is “better.” They are simply related, the way two singers in the same choir are related: different voices, same song.

Now plot the spread: on each day, the difference between the two, adjusted for how much of each stock you hold. What you get is a single line that wanders around a central level — its long-run average, or mean. When Stock A has a strong week and Stock B doesn’t, the line stretches upward. When the excitement fades, it drifts home. That wandering line is the spread chart, and it is the pairs trader’s entire world.

Standardize it, so “stretched” has a meaning

A raw spread is hard to judge. Is today’s gap big, or just Tuesday? The fix is to measure the gap in units of its own normal wiggle — standard deviations, written σ. Statisticians call the result a z-score:

  • z near 0 — the pair is at its normal distance. Nothing to do.
  • z near ±1σ — a mild lean. Still ordinary weather.
  • z near ±2.5σ — a genuine dislocation. The relationship is stretched well past its usual range.
  • z near ±3.25σ — a deep dislocation, the kind that shows up rarely.

Drawn on the chart, those levels become horizontal bands. The spread line spends most of its life inside the inner bands, and only occasionally visits the outer ones. Those rare visits are where the trades are.

Top: Stock A and Stock B, moving together, then apart, then back. Bottom: the same story told as one line — the spread in standard deviations. The trade exists only in the bottom panel. Generic illustrative prices, not a real pair.

Where the entry comes from

The rule used here is deliberately patient: no trade until the spread reaches ±2.5σ. Inside that band, the gap is noise — it might close tomorrow or wander further, and there is no edge in guessing. At ±2.5σ, the stretch itself is the information. One stock has been marked up or down out of proportion to its partner, and history says gaps like that usually narrow.

The position is the opposite of the stretch. If Stock A ran ahead, the trade is long Stock B and short Stock A, in matched size — a bet on the relationship, not on either company and not on the market. If the dislocation deepens to ±3.25σ, the system adds once, at the better price the deeper stretch provides. One add, defined in advance — not a ladder of hope on the way down.

Where the exit comes from — and why it’s early

The exit surprises people: the position closes when the spread recovers to ±1.0σ in absolute value, long before it reaches its mean. Three reasons. First, most of the healing happens in that first stretch of the journey home; the last part is the slowest and least reliable. Second, capital sitting in a nearly-healed pair is capital that can’t take the next genuine dislocation. Third, holding out for the perfect exit is how a statistical trade quietly becomes a story you tell yourself.

And the honest part: sometimes the gap isn’t a dislocation, it’s news. One company’s situation has genuinely changed, the old relationship is broken, and the spread is not coming home. That is what a pre-set hard stop is for. A spread chart can’t tell you why a gap opened — it can only tell you the gap is unusual. The rules respect both halves of that sentence.

How to look at a spread chart, in one pass

  • Find the mean — the level the line keeps coming home to.
  • Check the bands — where is the line now: ordinary weather, or out at ±2.5σ and beyond?
  • Ask how it got there — a sharp jump reads differently from a slow drift, even at the same level.
  • Remember what the chart can’t say — it shows the distance, never the reason.

That is the whole visual language of pairs trading: two stocks, one line, a handful of bands, and the discipline to act only at the edges. Everything else — selection, sizing, execution — is built on top of being able to read that one chart honestly.

Can I buy the flagship system?

Not yet. It is traded live by its developer, and it is not being offered for sale or licensing at this time. There are deliberately no purchase links on this page — when the product is ready, this is where that will be announced.

Where are the performance numbers?

No independently verified performance record is being offered for the retail system on this page. The live-trading statement is a description of the developer’s own activity, not a performance claim. The only return figures on this page are in the Performance section, and they are labelled hypothetical backtest results, raw without filtering, with their limits stated in full.

Is this investment advice?

No. CounterPoint publishes trading systems and research notes. Nothing here is a recommendation to buy or sell any security, and the illustrative charts use generic stocks precisely so they cannot be read as one.