Page views: every page opened. Unique visitors: different people, each counted once a day (a person who comes back tomorrow counts again). Counted anonymously by the site, with no IP addresses or IDs. Page views count up to 25 per person a day. Counting started October 12, 2026.

Targets are saved on this device. Weekly goals count the last 7 days.

This season's picks (NHL from the regular season on) with a market price, each graded as a flat $1 bet at that price. The likely range is where the true return probably sits; treat an edge as real only once the whole range is above 0%.

▲▼ next to a number: up or down since the previous day's results (hover one for the earlier value).

Every graded model parlay (NHL, NBA and NFL sets), each counted as a 1-unit bet at the legs’ prices multiplied. Hit rate is what happened; Model expected is the average chance the model gave them. Voids (every leg dropped) don’t count.

By payout band

Each cell: hits of graded, and profit in units. All: the three sets together. A 3-3.5x parlay needs to hit about 1 in 3 to break even, a 4-4.5x one about 1 in 4.

Up to three a day from 11 AM ET, NHL picks at Polymarket prices: one paying 3-3.5x, one 3.5-4x and one 4-4.5x, each 35-55% by the model, no player in two of them. Frozen once made and graded as a 1-unit bet at the legs’ prices multiplied (a book’s same-game parlay pays less). Team only.

Up to three a day from 11 AM ET, NBA picks at Polymarket prices, the same rules as the NHL set and its own record: one paying 3-3.5x, one 3.5-4x and one 4-4.5x, each 35-55% by the model, no player in two of them. Frozen once made and graded as a 1-unit bet at the legs’ prices multiplied (a book’s same-game parlay pays less). Team only.

Up to three a day from 11 AM ET, NFL picks at Polymarket prices, the same rules as the NHL set and its own record (tracking only; most days have no NFL games): one paying 3-3.5x, one 3.5-4x and one 4-4.5x, each 35-55% by the model, no player in two of them. Frozen once made and graded as a 1-unit bet at the legs’ prices multiplied (a book’s same-game parlay pays less). Team only.

One a day from 11 AM ET, 2-4 legs from any sport at Polymarket prices, paying 3x or more and over 50% by the model. No NHL assists (Polymarket won’t parlay them); NBA unders only when Polymarket sells them with $25+ offered, like every leg; and never a player from today’s 3 Picks of the Day (those come first). Frozen once made (Lock it below and no update can remake it) and graded as a 1-unit bet at the legs’ prices multiplied (a book’s same-game parlay pays less). Some days nothing fits: over 50% at 3x+ means the model rates it at least 1.5 times the market. Posted on Instagram as Parlay Payday (day 1: Oct 28, 2026; the record here starts that day). Team only.

Full-game spreads on Polymarket US in every sport, picked by a team-strength model built from final scores (spreads.py). A pick when the model is 5-15 points above a price of 35-65¢ with $25+ offered, at most one a game, from 11 AM ET; frozen once made and graded from the final score (overtime counts). Model expected is the average chance the model gave the picks.

Today

Three picks a day from the Secret methods, made from 11 AM ET and never changed after: fitting a method that's in profit in that sport (30+ graded picks there), priced 35-75¢ (pays about 1.33x-2.9x), the model at 60%+ and 5+ points over the price, no injury news, a price that hasn't dropped, three different games. The script and caption never name a method. Once you've filmed, press Lock today's picks: the updates then keep that day's picks exactly as they are (needs GitHub connected, as in Expenses).

Record

How each private method performs, graded by the edge tracker every update: every pick that fit it, each a 1-unit bet at its price. Each row counts every past pick that fits the method, including the ones it was found on. Under the name: the day it was added and its record since then, the live test of whether it really works (the number buyers would have seen). Likely range: where the true return probably is; a range above 0 is an edge holding up.

Every pick on today's slate (ET) that fits each method, followed live from ESPN box scores: these are the picks the edge tracker will grade for it, each a 1-unit bet at its price. A pick's price can still move until it locks shortly before its game, so a pick can join or leave a method until then. Picks reload every 5 minutes; live numbers every 30 seconds.

Rules the edge tracker grades without them being in the public code. They're saved encrypted with the team password and show up in the tracker from the next update. Empty fields mean any. Saving uses the GitHub connection from the Expenses tab.

• Loading…

One tap: the latest results image and today's caption with its tracked link. On a phone, Share opens Snapchat, Instagram or any app with both.

Written from today's picks and the latest results, each with a tracked link for its app, so its visits show up under Tracked links.

Every day after 11 AM ET the bot puts three short videos and one photo carousel in the @propstreaklab TikTok inbox as drafts (tiktok_post.py): the sports playing that day, rotating between value spots, hit streaks and the season record. Open TikTok, add a sound, paste the caption and post. Post or delete them the same day: TikTok holds about 5 drafts at a time.

Everyone who arrives through an affiliate's link, over the last 30 days. Average time on site is how long the site stayed on their screen during that visit, across the home and sport pages they opened (player and game pages, and time in another tab, don't count). It's an estimate: the site notes when a visit passes 10s, 20s, 1, 3, 10 and 30 minutes, and a visit that ends between two of those counts as halfway; 30 minutes or more counts as 30. Stayed 20s+ is the agreement's bar for a new viewer, measured here for every visit, new or returning. Time on site started being measured on Oct 13, 2026, so links from before then show fewer timed visits than visits.

Each row needs a written warning by email (agreement sections 4 and 6). Open email starts it in your mail app with the right warning filled in, or Copy email copies the same text. Press Mark warning sent once it's gone out, so it counts toward the two warnings and nobody gets it twice. A third breach gets the ending notice instead.

Makes a story-sized image of the latest results for each sport, ready for Snapchat or Instagram. Add your tagged link from the Link builder as a link sticker.

Makes a ready-to-upload ad from today's live numbers, with the text to paste into Ads Manager and a tracked link, so ad visits show up under Facebook in Tracked links. Feed is 1080 × 1080, Story / Reel is 1080 × 1920. Video is a 15-second Reel cut from a clip you film: pick it under Size.

Example ads

The first six concepts, made Oct 11 with that day's numbers. C1 is a one-day example; make a fresh one above.

Download PNG

Download PNG

Download PNG

Download PNG

Download PNG

Download PNG

Before you run them

• Meta treats real-money gambling ads as restricted. Keep ads about research and the public record: no "bet", "lock", "free money" or exchange names, and say "18+" on the ad (every ad here does).
• Target 18+ only (the site's age check), and decide which states to leave out before the first campaign.
• Hit rates need their price. Top 25 picks hit often because they're priced 85¢+. Keep the units line on record ads, and never write "guaranteed".
• The landing page shows the Kalshi / Polymarket sign-up pills. Meta's review may count those as gambling promotion. Consider hiding them for visitors arriving from Facebook.
• Video ads follow the same rules: the script avoids those words, and the end card carries the 18+ line. Read the script as written, or keep any changes to research and the record.
• Start small: 3–4 ad kinds, 2–3 versions each, about $20–30 a day per ad set. After 5 days, cut the weakest and make more of the best.