[Figure: ResultsTldr] ## Overview Front-of-center is one of the most debated topics in all of archery. This test is our attempt to separate fact from fiction on the topic. Due to the nature of arrow construction, we cannot change front-of-center while holding every other variable constant. For example, stacking front weight changes front-of-center, total arrow weight, dynamic spine, and arrow speed all at once. For this test, we built a matrix of 34 arrow builds spanning from roughly 10% to 25% front-of-center by combining four nominal spine values (200, 250, 300, 340), two Easton shaft families (5.0 and FMJ Max), and a range of internal FACT weight behind the same point. We then ran the same test protocol on each build, and ran a regression model to tease out which variables lead to which performance improvements. This article is an overview of our findings. ## Test Methods Every build in this article was shot from the same Hoyt AX3 33 (28″ draw, 70# draw weight) out of the Easton Precision Shooting Machine at 70 yards. In the days leading up to the test, each individual arrow was bare shaft nock tuned through paper. On test day, each build was then paper-tuned to a bullet hole at 15ft with the Hoyt XTS system before any groups were shot, off a fixed centershot. The only thing that changed across the 34 arrows was the shaft and the total point weight. For full details on the test methods, build matrix, and analysis processes, check out the [Front-of-Center Testing Overview](/research/arrow-study-2026/methods/foc/) and [Front-of-Center Analysis Overview](/research/arrow-study-2026/methods/foc-analysis-overview/) pages. ## Measures of Performance These are the primary measures we used to evaluate how a build performed: - **Broadhead Group Tightness from a Tuned Bow:** mean radius of a 6-arrow QAD Exodus group at 70 yards, with the bow paper-tuned to a bullet hole at 15ft. Lower is better. - **Broadhead Group Tightness from a Torqued Bow:** same metric, with a calibrated lateral torque on the riser to simulate shooter grip error. Lower is better. - **Broadhead Drift past Field Point:** with the bow torqued the same way, how far the broadhead landed from where the field point landed at 70 yards. This is an approximation of forgiveness to shooter error. Lower is better. - **Broadhead Drift from Synthetic Aim:** when we torqued the bow 2 degrees, the aim point at 70 yards shifted by 18in. The same torque induced a tail-right paper tear, causing the arrow to drift left. This is a similar but more understandable version of drift past field point. Lower is better. ## How a Regression Works This test protocol and analysis method is more complicated than our other tests. One key concept to understand is what a regression does and how we used one here. Imagine you want to figure out how much a single bedroom adds to the value of a house. You could plot price against number of bedrooms across a bunch of houses and read the slope. The problem is that houses with more bedrooms also tend to be bigger overall, sit in different neighborhoods, and have different square footage. A simple bedrooms-vs-price scatter can't tell you whether price rose because of the bedroom or because of all the other things that came along with it. A regression is a tool that uses all the data at once to estimate what each input is contributing on its own. The answer it gives is: "holding the other things equal, what does one more bedroom add to the price?" Same idea here. We can't just plot front-of-center against broadhead group size and call it done, because builds with high front-of-center in our matrix also tended to be heavier, have different static spine, and be slightly slower. The regression lets us ask: "holding the other build details equal, what does each percentage point of front-of-center do to broadhead group size?" The full math is on the [Front-of-Center Analysis Overview](/research/arrow-study-2026/methods/foc-analysis-overview/) page if you want it. For this article, just hang on to the idea that the numbers we report below are "the effect of each input on its own, with the other build details held equal," not what you'd see in a raw scatter. ## The Variables We Looked At We ran the regression to isolate two main things. - **Isolated front-of-center.** What does each percentage point of front-of-center do to each of our four measurements, with the other build details held equal? - **Isolated dynamic spine.** Dynamic spine is shorthand for how stiffly the arrow flexes during the shot. It's mostly determined by two things: the static spine of the shaft (how stiff the shaft is when bench-tested) and the total point weight (the weight sitting at the front of the arrow). A stiffer static spine and a lighter total point weight both contribute to a stiffer dynamic spine. We don't measure dynamic spine directly; we look at static spine and total point weight separately in the regression and report them side by side. ## The Isolated Front-of-Center Effect The chart below shows the isolated front-of-center effect on each of the four measurements. Each row is one measurement. The dot is the predicted change at +5 percentage points of front-of-center (so, going from 13% to 18%, for example), translated to inches at 70 yards. The bar around the dot is how sure we are about that estimate. Negative numbers (left of zero) are good on every row, because tighter groups, less drift past field point, and less drift from synthetic aim are all what you want. A tight bar clearly to the left of zero is a confident improvement. A wide bar that crosses zero is closer to "the data can't tell." On the two group-tightness measurements, the isolated front-of-center effect is clearly negative. More front-of-center, tighter groups. +5 percentage points of front-of-center predicts roughly 2 inches tighter mean radius on the tuned bow at 70 yards, and about 1 inch tighter on the torqued bow. On the two drift measurements (forgiveness to shooter error), the bars are wider and cross zero. The reason is discussed below in the "A Note on Forgiveness to Shooter Error" section. [Figure: PlotSections] Predicted change at +5 percentage points of front-of-center, in inches at 70 yards, with total arrow weight and the other build details held equal. Bars are 95% confidence intervals; lower (left of zero) is better. Solid dots are results we're confident about; hollow dots are ones where the bar still crosses zero (the data can't tell). ## The Isolated Dynamic-Spine Effect The chart below shows the isolated effect of the two components of dynamic spine on each of the four measurements: static spine (a stiffer shaft) and total point weight (a lighter front). Both components moved broadhead groups in the stiffer-dynamic-spine direction. Stepping 100 nominal spine units stiffer (for example, a 300 shaft to a 200) predicts about 0.9 inches tighter mean radius on the tuned bow broadhead group at 70 yards. Dropping 50 grains of total point weight, with front-of-center already in the model, predicts about 0.7 inches tighter on the same measurement. Smaller than the front-of-center effect, but both point the same way: a stiffer dynamic spine helped. As a sanity check on the spine result, we also ran the regression with a hand-measured static spine instead of the manufacturer label, and the results were the same. [Figure: PlotSections] Predicted change for two stiffer-dynamic-spine moves: a static-spine step of +{NOMINAL_SPINE_STEP_UNITS} nominal spine units stiffer (e.g. a 300 shaft to a 200) and a {POINT_WEIGHT_STEP_GR}-grain drop in total point weight. Front-of-center is already in the regression, so each row isolates the component on its own. Bars are 95% confidence intervals; lower (left of zero) is better. Solid dots are results we're confident about; hollow dots are ones where the bar still crosses zero (the data can't tell). ## A Note on Forgiveness to Shooter Error The two drift measurements (drift past field point and drift from synthetic aim) are different from the group-tightness measurements in one important way. The answer to "did front-of-center help?" depends on whether we control for total arrow weight and launch speed in the regression. If we don't control for arrow weight and launch speed, more front-of-center clearly predicted less drift on both measurements. +5 percentage points of front-of-center predicts about 1 inch less broadhead drift past field point, and a bit over 2 inches less broadhead drift from synthetic aim, at 70 yards. The chart below shows those estimates with their confidence intervals. [Figure: PlotSections] Predicted change at +5 percentage points of front-of-center, in inches at 70 yards, without holding total arrow weight or launch speed equal. This is the "whole-build" view, where higher-FoC arrows were also a little heavier and slower. Bars are 95% confidence intervals; lower (left of zero) is better. Solid dots are results we're confident about; hollow dots are ones where the bar still crosses zero (the data can't tell). ## Weight, Speed, and Which Reading Applies We feel reasonably comfortable treating launch speed as part of the package rather than a variable to control for, because a separate 2026 test of ours compared the same vanes and broadheads at ~290fps and ~325fps and didn't find a meaningful accuracy or forgiveness difference between the two. See [Are Fast Arrows Less Forgiving?](/research/arrow-study-2026/results/speed-vs-accuracy/) for the full comparison. If we hold total arrow weight constant (asking what front-of-center does on its own), the data isn't sure. The bar crosses zero on both drift measurements in the isolated chart above. In our 34 builds, the high front-of-center arrows tended to be heavier and slightly slower, because the front-of-center came from stacking inserts. The regression can't cleanly separate "more front-of-center" from "more weight" on the drift measurements when those moved together. This is also a sample-size story. With 34 builds, the regression has limited power to untangle predictors that moved together in our matrix. A larger build matrix, or one that crossed point weight and front-of-center more independently, would give a sharper read on which version of the answer is right. ## Takeaway In our 34 arrows, three things improved performance, and we can't move them independently on a real arrow. At the same time, we want to: 1. Increase front-of-center 2. Increase static spine stiffness 3. Decrease total point weight For example, adding total point weight to raise front-of-center is the obvious move. However, this hurts performance by the other two. Increasing stiffness usually also means an increase in shaft weight, which hurts front-of-center. There are also other factors we did not consider here: - Adding point weight and stiffness also means an arrow that weighs more overall, which hurts other aspects of the setup like ranging-error forgiveness (how far you can be off on your range to the animal and still catch vitals). - Lighter GPI shafts generally aren't as durable. The practical reading isn't "always go stiffer" or "always increase point weight." It's to raise front-of-center efficiently. Look for moves that pull the balance point forward without piling more mass behind the same point. A slightly stiffer shaft that stays inside what the bow can be tuned to. A shaft with a lower GPI (mass per inch) so more of the weight budget can sit in front of the balance point. A few caveats: - All of our front weight lived inside the shaft as FACT inserts. A heavier external broadhead places the same mass differently and may flex the shaft differently. We can't speak to that from this data. - Insert mass, insert length, and insert balance point all moved together across our arrows. We can't say whether insert mass alone is the thing that matters or whether insert geometry plays an independent role. - "Stiffer better" in our data is stiffer-better inside the 200 to 340 nominal spine range we tested, on the bow we used. Over-stiffening relative to what your bow can tune to isn't something we measured. The XTS system was easily tuned to all of these arrow builds. - One bow, one draw weight, one draw length. The exact inch numbers may not transport unchanged to a different setup. - Each build was scored on one 6-arrow group per condition. With six arrows, any single build's mean radius carries some random scatter around its true value, which adds to the width of the bars in our charts. - Outside the tested range (roughly 10% to 25% front-of-center, nominal spines 200 to 340), we don't have evidence either way.