ANY estimate based on square footage of living space is lousy. Square feet of floor area mean NOTHING! It's all about the square feet & U-value of
exterior wall & roof area, and the amount of air-leakage into the structure, which vary a LOT for the same amount of floor space- heat loads per ft^2 can & will vary as much as 300%, or more in the same climate, similar sized house.
Manual-J often misses the mark to the high-side by as much as 50% (which is way better than 300%, but still not great). Knowing the average BTU/ft^2/HDD for homes in your area may be a useful starting point to see if your fuel consumption or efficiency (heating system or building-envelope) is off the charts, but it's clearly the wrong way to design a heating system that's both efficient and delivers the heat even when it's cold out.]
But doing what you're doing, MEASURING your fuel use per HDD is better than manual-J for calculating your as-used heating load. (Forget the per-square foot part, which isn't irrelevant.) By knowing (or looking up) your burner output, and fuel use per HDD, you can then get to within 10% of the your true heat load. If based on your burner output, system type you're running less than 100% duty cycle on the burner at "design day" temps, you have to derate the efficiency a bit and re-calc based on regression curves for oversized systems.
Here's how to do it:
Let's say your 97th percentile lowest binned hourly temperature (the "design day" temperature) is 0F. If you use 65F as your base heating degree-day temp (better insulated houses can use 60F), Then your design heat output need to support the difference, 65F-0F=
65F.
IIRC, ACCA uses the 97th percentile hourly-temperature bin is usually the recommended design-temp number, and ASHRAE uses 99th, but even 95th percentile would be generally "safe" in most instances. Temps below the 95th percentile usually don't persist for more than a few hours, and they're usually at night when you're all cozy in bed. To estimate what your design-temp should be, look up the coldest temp your area saw in the past few years, add 5F, and you'll be "in the range" (unless it was a 25 or hundred year all-time low or something.)
If by looking at the published heating degree days for a particular billing period (there are weather-history data sources online-pick a city near you) you can calculate your fuel use per HDD: Let's say you used 100gallons of oil for a period that added up to 500 total degree days, is 100/500= 0.2gallons/degree day.
To convert that to BTUs for heating oil its 138000BTUs/gallon so that's 138000 x 0.2=
27600 BTUs/HDD fuel use.
Assuming your furnace/boiler has an efficiency rating of 80%, the actual heat delivered to the heating distribution (ducts or radiators, etc) is 0.8 x 27600=
22080 BTUs/HDD output.
Now, a heating degree-DAY is 24 hours, and heating equipment is rated in output per hour so you have to do more arithmetic. The 65F design temp delta would be 65HDD if it stayed that cold all day, so your hourly output need would be (65 x 22080)/24=
59800BTUs/hour.
If that's the rated output of your current system, congratulations- you just won the lottery- it's PERFECTLY sized.
But as is more typical, let's say the burner rating is 225000 BTU/hour in- 180000BTU/hour out (80% efficiency), that means it is at least 3x oversized. That's where the regression curves come in for a closer/better guesstimate. Take a look at figure 1:
http://simulationresearch.lbl.gov/dirpubs/42175.pdf
If you have a 180K output and at most a 60K load, you're at the 0.33 load mark for DESIGN DAY load (for a few hours on that day). Most of the heating season you're in the 0.11-0.22 fractional load range (1/3-2/3 of of your design-day load.) So, if you have a hot-air furnace or very low mass boiler, the SDL-C111 curve, which looking at the curve, you need to derate your efficiency estimate to ~0.8x the stated thermal or AFUE efficiency of the equipment: 0.8x 80%=
64% -as-used efficiency. (If you have a cast-iron hydronic boiler other high-mass system or a steam boiler use the other curves. Just take a guess at which one, it'll be close enough.)
Then using that as your newer-better estimate your actual design-day heat load based on derated estimate is:
0.8 x 59800 BTU/hour= 47840 BTU/hour.
THAT's your real as-used heat load to within 10%. (I'ts actually slightly to the high side on average, since at a ~47.8K load a 180K burner is~3.7x oversized, not ~3x, but the regression curves aren't perfectly matched to YOUR equipment.). So the 47800 BTU/hour output is the number to use when sizing the equipment for maximum annual efficiency. If you keep it to no more than 15% over that you'll likely beat AFUE numbers for efficiency. Alternatively, if you go very much below that, count on needing some supplemental heat for some the coldest of cold nights (got a decent small wood stove?)
And that's WAY more accurate than Manual-J (poor-man's or rich.) It's a snapshot of how well your insulation etc is performing, and how you're actually utilizing the heating system. Manual-J numbers often run 25-35% over real-world measurements like this, unless you're the type who keeps it at 72F day & night. Going more than 10% over Manual-J you're bound to see a measurable difference in fuel use. But oversizing it by 25% from your measurement won't kill it on efficiency- look at the 75% then the 25-50% range on the curves- it's not ultra-derated (but some.) Still, if you're planning on doing another round of envelope-upgrades (new windows, air-sealing, insulating, etc.) you could soon find yourself 2x+ oversized and slipping over the efficiency cliff on the left edge of regression curves again if you oversize it much based on your measurements.
Oversizing a heating system typically costs you more up front, and for every year thereafter. It's better to undersize it a bit, spend the difference on envelope upgrades, and make your annual-payments on upgrading the building envelope rather than burning the extra fuel.