diff --git a/bin.src/plot_astrometric_residuals-from_pickle.py b/bin.src/plot_astrometric_residuals-from_pickle.py
new file mode 100644
index 00000000..8cda1030
--- /dev/null
+++ b/bin.src/plot_astrometric_residuals-from_pickle.py
@@ -0,0 +1,77 @@
+#!/usr/bin/env python
+"""Plot the pickles produced by `plot_astrometric_residuals.py`
+
+Originally written to tweak the plots for John Parejko's AAS 233 poster.
+
+Does not depend on any lsst code: pass it around with the pickle files
+produced by `plot_astrometric_residuals.py` to experiment with various
+plotting options and libraries.
+"""
+import os
+import pickle
+
+import numpy as np
+
+import matplotlib.pyplot as plt
+import seaborn
+seaborn.set_style('white')  # noqa: E402
+seaborn.set_context("poster")  # noqa: E402
+
+
+def uv_mean(xlim, ylim, xx, yy, uu, vv):
+    """Compute the mean of uu and vv on a grid of 0,width x 0,height."""
+    nx = 30
+    ny = 60
+    uMean = np.zeros((nx-1, ny-1))
+    vMean = np.zeros((nx-1, ny-1))
+    xMean = np.zeros((nx-1, ny-1))
+    yMean = np.zeros((nx-1, ny-1))
+    ww = np.linspace(xlim[0], xlim[1], nx)
+    hh = np.linspace(ylim[0], ylim[1], ny)
+    for i, (w0, w1) in enumerate(zip(ww[:-1], ww[1:])):
+        inx = (xx >= w0) & (xx <= w1)
+        for j, (h0, h1) in enumerate(zip(hh[:-1], hh[1:])):
+            iny = (yy >= h0) & (yy <= h1)
+            inside = inx & iny
+            xMean[i, j] = (w0 + w1)/2
+            yMean[i, j] = (h0 + h1)/2
+            uMean[i, j] = np.mean(uu[inside])
+            vMean[i, j] = np.mean(vv[inside])
+
+    return xMean, yMean, uMean, vMean
+
+
+def plot_quiver(xx, yy, uu, vv, ccd, label):
+    """Make a quiver plot of the astrometry error vectors."""
+    cycle = plt.rcParams['axes.prop_cycle'].by_key()['color']
+
+    color = cycle[0] if 'jointcal' in label else cycle[1]
+    scale = 0.25 if 'mean' in label else 1
+
+    fig = plt.figure(figsize=(6, 10))
+    ax = fig.add_subplot(111)
+
+    Q = ax.quiver(xx, yy, uu, vv, units='x', pivot='tail', scale=scale, width=7,
+                  headwidth=4, clip_on=False, color=color)
+    length = 5/scale if 'mean' in label else 100
+    key = ax.quiverkey(Q, 0.85, 0.90, length, '%s mas'%(length), angle=45,
+                       coordinates='figure', labelpos='W', fontproperties={'size': 24})
+
+    ax.set_xticklabels([])
+    ax.set_yticklabels([])
+
+    filename = "plots/quiver-%s-%s.png"%(ccd, label)
+    plt.savefig(filename, bbox_extra_artists=(key,))
+    plt.close(fig)
+
+
+# NOTE: swap this to change whether to read and plot processCcd or jointcal output.
+name = 'single'
+# name = 'jointcal'
+# the path where the pickle/ directory is
+path = '~/lsst/temp/AAS2019/quiver'
+filename = os.path.join(path, 'pickle/quiverData-%s-6.pickle'%name)
+with open(os.path.expanduser(filename), 'rb') as infile:
+    xx, yy, uu, vv, xlim, ylim, ccd = pickle.load(infile)
+    xMean, yMean, uMean, vMean = uv_mean(xlim, ylim, xx, yy, uu, vv)
+    plot_quiver(xMean, yMean, uMean, vMean, ccd, 'mean-'+name)
diff --git a/bin.src/plot_astrometric_residuals.py b/bin.src/plot_astrometric_residuals.py
new file mode 100644
index 00000000..f0d089bb
--- /dev/null
+++ b/bin.src/plot_astrometric_residuals.py
@@ -0,0 +1,293 @@
+#!/Users/parejkoj/lsst/lsstsw/miniconda/envs/lsst-scipipe/bin/python  # noqa
+"""Plot astrometric residuals for ccds in a tract.
+
+Writes output to the current working directory, in `plots/` and `pickles/`,
+which are created if they do not exist.
+
+Originally written to make plots for John Parejko's AAS 233 poster.
+
+Example to run on tract 8524 of HSC PDR1 on lsst-dev:
+
+TRACT=8524
+VISITS=7286^7288^7298^7300^7302^7304^7310^7338^7340^7350^7352^7356^7358^7364^7366^7370^7372^7378^7384^7386^7390^7392^7394^7396^7400^7402^7416^14124^14126^14128^14130^14142^14144^14146^14164^14166^14176^14178^14196^14198^14206^14208^14210^1623
+CCDS=0..8^10..103
+DATADIR=/project/parejkoj/hscRerun/DM-15713/WIDE
+plot_astrometric_residuals.py $DATADIR --jointcal --id ccd=$CCDS visit=$VISITS tract=$TRACT
+
+Or to run on jointcal's cfht testdata output in your local directory:
+
+DATADIR=/Users/parejkoj/lsst/jointcal/jointcal/tests/.test/JointcalTestCFHT/test_jointcalTask_2_visits_constrainedAstrometry_no_photometry
+TRACT=0
+VISITS=849375^850587
+CCDS=12^13^14^21^22^23
+plot_astrometric_residuals.py $DATADIR --jointcal --id ccd=$CCDS visit=$VISITS tract=$TRACT
+"""
+
+import collections
+import pickle
+import os
+
+import numpy as np
+import astropy.units as u
+
+import matplotlib.pyplot as plt
+import seaborn
+seaborn.set_style('white')  # noqa: E402
+seaborn.set_context("poster")  # noqa: E402
+
+from lsst.afw.cameraGeom import PIXELS, FOCAL_PLANE
+import lsst.afw.table
+import lsst.daf.persistence
+import lsst.geom
+import lsst.meas.astrom
+import lsst.pex.config
+import lsst.pipe.base
+from lsst.validate.drp.util import positionRmsFromCat, averageRaDecFromCat
+
+
+def filter_matches(allMatches, fluxField):
+    """Filter down to matches with at least 2 sources and good flags."""
+    flagKeys = [allMatches.schema.find("base_PixelFlags_flag_%s" % flag).key
+                for flag in ("saturated", "cr", "bad", "edge")]
+    nMatchesRequired = 2
+
+    fluxSnrKey = allMatches.schema.find(fluxField + "_snr").key
+    # NOTE: alias oddities mean I have to change "_flux" <-> "_instFlux" depending on catalog version.
+    fluxKey = allMatches.schema.find(fluxField + "_flux").key
+
+    def goodFilter(cat, goodSnr=10):
+        if len(cat) < nMatchesRequired:
+            return False
+        for flagKey in flagKeys:
+            if cat.get(flagKey).any():
+                return False
+        if not (cat.get(fluxKey) > 0).all():
+            return False
+        snr = np.median(cat.get(fluxSnrKey))
+        # Note that this also implicitly checks for snr being non-nan.
+        return snr >= goodSnr
+
+    return allMatches.where(goodFilter)
+
+
+def prep_matching(butler, dataRef):
+    """Prep MultiMatch with the catalog schema and field types."""
+    oldSchema = dataRef.get('src_schema').schema
+    fluxField = oldSchema.getAliasMap().get("slot_CalibFlux")
+
+    # make the new schema, with a field for S/N
+    mapper = lsst.afw.table.SchemaMapper(oldSchema)
+    mapper.addMinimalSchema(oldSchema)
+    mapper.addOutputField(lsst.afw.table.Field[float](fluxField + '_snr', 'flux SNR'))
+    newSchema = mapper.getOutputSchema()
+    newSchema.setAliasMap(oldSchema.getAliasMap())
+
+    types = {"visit": np.int32, "ccd": np.int32}
+    multiMatch = lsst.afw.table.MultiMatch(newSchema, types)
+    return fluxField, newSchema, mapper, multiMatch
+
+
+def read_one(dataRef, useJointcal=False):
+    # NOTE: I know we're not supposed to access the internals of dataRefs like this,
+    # but I can't be bothered to do the "getDetector()" stuff, and you can't
+    # get at a visit ID in any other way anyway...
+    visit = dataRef.dataId['visit']
+    ccd = dataRef.dataId['ccd']
+    try:
+        oldCat = dataRef.get('src')
+    except lsst.daf.persistence.butlerExceptions.NoResults:
+        # ignore missing data
+        print("No data:", visit, ccd)
+        return None
+    if useJointcal:
+        wcs = dataRef.get('jointcal_wcs')
+        lsst.afw.table.updateSourceCoords(wcs, oldCat)
+
+    return visit, ccd, oldCat
+
+
+def do_match(multiMatch, butler, dataRefs, fluxField, newSchema, mapper, useJointcal=False):
+    """Make the multiMatch, identify good matches, and compute aggregate statistics."""
+    # import concurrent.futures
+    # import itertools
+    # max_workers = 1
+    # with concurrent.futures.ThreadPoolExecutor(max_workers=max_workers) as executor:
+    #     mapped = executor.map(read_one, dataRefs, itertools.repeat(useJointcal))
+
+    for dataRef in dataRefs:
+        # for (visit, ccd, oldCat), dataRef in zip(mapped, dataRefs):
+        visit, ccd, oldCat = read_one(dataRef, useJointcal)
+        catalog = lsst.afw.table.SourceCatalog(newSchema)
+        tmpCat = lsst.afw.table.SourceCatalog(lsst.afw.table.SourceCatalog(newSchema).table)
+        tmpCat.extend(oldCat, mapper=mapper)
+        tmpCat[fluxField + '_snr'][:] = tmpCat[fluxField + '_instFlux'] / tmpCat[fluxField + '_instFluxErr']
+        catalog.extend(tmpCat, False)
+
+        multiMatch.add(catalog, dataId=dataRef.dataId)
+        print("Loaded:", visit, ccd, len(tmpCat))
+
+    matchCat = multiMatch.finish()
+    allMatches = lsst.afw.table.GroupView.build(matchCat)
+    print("Found matches, groups:", len(matchCat), len(allMatches))
+
+    goodMatches = filter_matches(allMatches, fluxField)
+    print("Good groups:", len(goodMatches))
+
+    averageCoord = goodMatches.aggregate(averageRaDecFromCat,
+                                         dtype=[('ra', np.float64), ('dec', np.float64)])
+    distance = goodMatches.aggregate(positionRmsFromCat) * u.milliarcsecond
+    return goodMatches, averageCoord, distance
+
+
+def count_ccds(goodMatches):
+    """Count how many objects are on each ccd."""
+    counts = collections.defaultdict(int)
+    for group in goodMatches.groups:
+        for x in group:
+            counts[x['ccd']] += 1
+    print("ccd counts:", ', '.join("%s: %s"%(k, v) for k, v in counts.items()))
+    return counts
+
+
+def compute_errors(counts, goodMatches, averageCoord, detector, focalplane=False):
+    """Return the ra/dec error for each centroided object.
+
+    If focalplane is set, return x,y in focal plane coordinates, instead of pixels.
+    """
+    detectorId = detector.getId()
+    if focalplane:
+        pixelToFocal = detector.getTransform(PIXELS, FOCAL_PLANE)
+
+    xx = np.zeros(counts[detectorId])
+    yy = np.zeros(counts[detectorId])
+    uu = np.zeros(counts[detectorId])
+    vv = np.zeros(counts[detectorId])
+    i = 0
+
+    for group, coord in zip(goodMatches.groups, averageCoord):
+        good = group['ccd'] == detectorId
+        n = good.sum()
+        if focalplane:
+            pixels = [lsst.geom.Point2D(x, y) for x, y in zip(group[good].getX(), group[good].getY())]
+            focals = pixelToFocal.applyForward(pixels)
+            xx[i:i + n] = [f.getX() for f in focals]
+            yy[i:i + n] = [f.getY() for f in focals]
+        else:
+            xx[i:i + n] = group[good].getX()
+            yy[i:i + n] = group[good].getY()
+        uu[i:i + n] = group[good]['coord_ra'] - coord[0]
+        vv[i:i + n] = group[good]['coord_dec'] - coord[1]
+        i += n
+    uu = (uu*u.radian).to_value(u.milliarcsecond)
+    vv = (vv*u.radian).to_value(u.milliarcsecond)
+
+    bbox = detector.getBBox()
+    if focalplane:
+        # convert the detector bbox to focal plane coordinates
+        focalBBox = lsst.geom.Box2D()
+        for point in pixelToFocal.applyForward(lsst.geom.Box2D(bbox).getCorners()):
+            focalBBox.include(point)
+        bbox = focalBBox
+
+    return xx, yy, uu, vv, bbox
+
+
+def uv_mean(xlim, ylim, xx, yy, uu, vv):
+    """Compute the mean of uu and vv on a grid within bbox."""
+    nx = 20
+    ny = 40
+    uMean = np.zeros((nx-1, ny-1))
+    vMean = np.zeros((nx-1, ny-1))
+    xMean = np.zeros((nx-1, ny-1))
+    yMean = np.zeros((nx-1, ny-1))
+    ww = np.linspace(xlim[0], xlim[1], nx)
+    hh = np.linspace(ylim[0], ylim[1], ny)
+    for i, (w0, w1) in enumerate(zip(ww[:-1], ww[1:])):
+        inx = (xx >= w0) & (xx <= w1)
+        for j, (h0, h1) in enumerate(zip(hh[:-1], hh[1:])):
+            iny = (yy >= h0) & (yy <= h1)
+            inside = inx & iny
+            xMean[i, j] = (w0 + w1)/2
+            yMean[i, j] = (h0 + h1)/2
+            uMean[i, j] = np.mean(uu[inside])
+            vMean[i, j] = np.mean(vv[inside])
+
+    return xMean, yMean, uMean, vMean
+
+
+def plot_quiver(xx, yy, uu, vv, ccd, label, focal):
+    """Make a quiver plot of the astrometry error vectors."""
+    cycle = plt.rcParams['axes.prop_cycle'].by_key()['color']
+
+    color = cycle[0] if 'jointcal' in label else cycle[1]
+    scale = 0.25 if 'mean' in label else 1
+    width = 7.0
+    if focal:
+        width = .1
+        scale *= 70
+
+    fig = plt.figure(figsize=(6, 10))
+    ax = fig.add_subplot(111)
+
+    # units of the vectors are not, in fact "x", but it's simpler to treat them that way.
+    # uu, vv should be in arcsec, xx,yy in pixels or mm (if focal)
+    Q = ax.quiver(xx, yy, uu, vv, units='x', pivot='tail', scale=scale, width=width,
+                  headwidth=4, clip_on=False, color=color)
+    length = 5/scale if 'mean' in label else 100
+    key = ax.quiverkey(Q, 0.85, 0.90, length, '%s mas'%(length), angle=45,
+                       coordinates='figure', labelpos='W', fontproperties={'size': 24})
+
+    ax.set_xticklabels([])
+    ax.set_yticklabels([])
+
+    plt.title('{}'.format(ccd))
+    filename = "plots/quiver-%s-%s.png"%(ccd, label)
+    plt.savefig(filename, bbox_extra_artists=(key,))
+    plt.close(fig)
+
+
+def main():
+    parser = lsst.pipe.base.InputOnlyArgumentParser("plot_astrometric_residuals",
+                                                    description=__doc__)
+    parser.add_argument("--jointcal", action='store_true',
+                        help="Use the jointcal_wcs dataset to update each ccd's coordinates.")
+    parser.add_argument("--focalplane", action='store_true',
+                        help="Make plots and pickle files in focal plane coordinates.")
+    # We need to specify "jointcal_wcs" here so that the butler understands what a `tract` is.
+    parser.add_id_argument(name="--id", datasetType="jointcal_wcs",
+                           help="data IDs, e.g. --id visit=12345 ccd=1^2 tract=1234")
+
+    args = parser.parse_args(config=lsst.pex.config.Config())
+    name = 'jointcal' if args.jointcal else 'single'
+    butler = args.butler
+
+    if len(args.id.refList) == 0:
+        raise RuntimeError("No data to process! Check your id list and/or input data path.")
+    fluxField, newSchema, mapper, multiMatch = prep_matching(butler, args.id.refList[0])
+    goodMatches, averageCoord, distance = do_match(multiMatch, butler, args.id.refList,
+                                                   fluxField, newSchema, mapper, useJointcal=args.jointcal)
+    counts = count_ccds(goodMatches)
+
+    os.makedirs('plots', exist_ok=True)
+    os.makedirs('pickles', exist_ok=True)
+
+    for ccd in counts.keys():
+        detector = butler.get('calexp_detector', ccd=ccd)
+        xx, yy, uu, vv, bbox = compute_errors(counts, goodMatches, averageCoord, detector,
+                                              focalplane=args.focalplane)
+
+        xlim = (bbox.getMinX(), bbox.getMaxX())
+        ylim = (bbox.getMinY(), bbox.getMaxY())
+        filename = "pickles/quiverData-%s-%s.pickle"%(name, ccd)
+        with open(filename, 'wb') as outfile:
+            pickle.dump((xx, yy, uu, vv, xlim, ylim, ccd),
+                        outfile,
+                        protocol=pickle.HIGHEST_PROTOCOL)
+
+        plot_quiver(xx, yy, uu, vv, ccd, 'all-'+name, args.focalplane)
+        xMean, yMean, uMean, vMean = uv_mean(xlim, ylim, xx, yy, uu, vv)
+        plot_quiver(xMean, yMean, uMean, vMean, ccd, 'mean-'+name, args.focalplane)
+
+
+if __name__ == "__main__":
+    main()