Daily Betting Insights: Thursday, January 29, 2026 | NHL & NBA Picks

Swept the card last night. 4-0. Columbus, Knicks, Charlotte, Cleveland. All green.

NHL overall went 2-1. Ottawa dumped Colorado, which most models missed, but Otterline AI actually had Ottawa. That's the in-house model, and it's been the top performer this month. The consensus got pulled toward Colorado by the other models, so that one stung. Columbus beat Philly 5-3, Islanders beat Rangers 5-2. Every winner scored five goals.

NBA went 5-4 overall but our filtered picks went 3-0. The Knicks were down early and came back hard. Chicago lost to Indiana by three. Close games all night.

Why NHL Needs Filters

Spent some time this episode breaking down the weekly charts. Here's what the data shows:

NBA consensus just works. Every week it's been above 50%. Started at 53%, then 52, 61, 61, 56, and this week it's at 74% (which will come down). The line stays steady. Verified value on top of that makes it even better.

NHL is different. Consensus actually underperforms individual models a lot of the time. It's noisy. Goalies steal games, refs make weird calls, one bounce changes everything. You can't just bet every consensus pick and expect to win.

That's where power plays comes in. When you filter for the strongest signals, the win rate jumps. Power plays hit 77% this week while consensus was at 47%. Last week: 69% vs 66%. The filter is what makes the model work in hockey.

If there's no good signal on a given night, we just don't bet. Yesterday had three games, we took one. That discipline matters.

Verified Juice

New strategy I'm using for picks. Instead of cherry-picking three games, I'm taking everything where we have at least 10% edge on the market.

Example: Vegas says Colorado has a 55% chance to win. Our model says 66%. That's 11% juice. We take it.

It removes human bias and keeps everything systematic. The picks below all clear that 10% threshold.

Today's Picks

Big slate. 15 NHL games, 8 NBA games. Filtered down to seven plays.

NHL

COL vs. MTL → COL 66% model confidence vs 56% market odds. 77 game sample size on this signal.

WSH vs. DET → DET

DAL vs. VGK → VGK

BUF vs. LAK → BUF

NBA

MIA → MIA 51% odds, 16% juice.

WAS → WAS 45% odds, 26% juice. Spicy underdog. All three models agree.

MIN → MIN (half unit) 33% odds. Heavy underdog so sizing down, but the signal is there.

Dissenter Tracking

New feature on the site. When one model disagrees with consensus, we now track how that dissenter performs.

Example: Tampa vs Winnipeg tonight. Three models say Tampa, MP says Winnipeg. When MP has disagreed, it's gone 12-7. That's noise. We skip it.

Another example: Florida vs St. Louis. Three models say Florida, Otterline says St. Louis. When Otterline has faded consensus, it's gone 7-5 against itself. That actually confirms our Florida pick.

This stuff doesn't show up in the headline record, but it helps separate clean signals from noisy ones.

Headlines

Trey Platt traded to the Islanders. Scored right away.

Kane hits US-born NHL points mark.

Dunkel model officially removed. It was dragging down the consensus. Should see improvement going forward.

Full slate and model breakdowns at otterline.club